Search

This research presents an integrated framework designed to automate the classification of pulmonary chest x-ray images. Leveraging convolutional neural networks (CNNs) with a focus on transformer architectures, the aim is to improve both the accuracy and efficiency of pulmonary chest x-ray image analysis. A central aspect of this approach involves utilizing pre-trained networks such as VGG16, ResNet50, and MobileNetV2 to create a feature ensemble. A notable innovation is the adoption of a stacked ensemble technique, which combines outputs from multiple pre-trained models to generate a comprehensive feature representation. In the feature ensemble approach, each image undergoes individual processing through the three pre-trained networks, and pooled images are extracted just before the flatten layer of each model. Consequently, three pooled images in 2D grayscale format are obtained for each original image. These pooled images serve as samples for creating 3D images resembling RGB images through stacking, intended for classifier input in subsequent analysis stages. By incorporating stacked pooling layers to facilitate feature ensemble, a broader range of features is utilized while effectively managing complexities associated with processing the augmented feature pool. Moreover, the study incorporates the Swin Transformer architecture, known for effectively capturing both local and global features. The Swin Transformer architecture is further optimized using the artificial hummingbird algorithm (AHA). By fine-tuning hyperparameters such as patch size, multi-layer perceptron (MLP) ratio, and channel numbers, the AHA optimization technique aims to maximize classification accuracy. The proposed integrated framework, featuring the AHA-optimized Swin Transformer classifier utilizing stacked features, is evaluated using three diverse chest x-ray datasets-VinDr-CXR, PediCXR, and MIMIC-CXR. The observed accuracies of 98.874%, 98.528%, and 98.958% respectively, underscore the robustness and generalizability of the developed model across various clinical scenarios and imaging conditions., (© 2024 IOP Publishing Ltd. All rights, including for text and data mining, AI training, and similar technologies, are reserved.)

Background and Objective: Accurate segmentation of brain tumors from multimodal magnetic resonance imaging (MRI) holds significant importance in clinical diagnosis and surgical intervention, while current deep learning methods cope with situations of multimodal MRI by an early fusion strategy that implicitly assumes that the modal relationships are linear, which tends to ignore the complementary information between modalities, negatively impacting the model's performance. Meanwhile, long-range relationships between voxels cannot be captured due to the localized character of the convolution procedure., Method: Aiming at this problem, we propose a multimodal segmentation network based on a late fusion strategy that employs multiple encoders and a decoder for the segmentation of brain tumors. Each encoder is specialized for processing distinct modalities. Notably, our framework includes a feature fusion module based on a 3D discrete wavelet transform aimed at extracting complementary features among the encoders. Additionally, a 3D global context-aware module was introduced to capture the long-range dependencies of tumor voxels at a high level of features. The decoder combines fused and global features to enhance the network's segmentation performance., Result: Our proposed model is experimented on the publicly available BraTS2018 and BraTS2021 datasets. The experimental results show competitiveness with state-of-the-art methods., Conclusion: The results demonstrate that our approach applies a novel concept for multimodal fusion within deep neural networks and delivers more accurate and promising brain tumor segmentation, with the potential to assist physicians in diagnosis., (© 2024 The Author(s). Journal of Applied Clinical Medical Physics published by Wiley Periodicals LLC on behalf of American Association of Physicists in Medicine.)

Neuron tracing, alternately referred to as neuron reconstruction, is the procedure for extracting the digital representation of the three-dimensional neuronal morphology from stacks of microscopic images. Achieving accurate neuron tracing is critical for profiling the neuroanatomical structure at single-cell level and analyzing the neuronal circuits and projections at whole-brain scale. However, the process often demands substantial human involvement and represents a nontrivial task. Conventional solutions towards neuron tracing often contend with challenges such as non-intuitive user interactions, suboptimal data generation throughput, and ambiguous visualization. In this paper, we introduce a novel method that leverages the power of extended reality (XR) for intuitive and progressive semi-automatic neuron tracing in real time. In our method, we have defined a set of interactors for controllable and efficient interactions for neuron tracing in an immersive environment. We have also developed a GPU-accelerated automatic tracing algorithm that can generate updated neuron reconstruction in real time. In addition, we have built a visualizer for fast and improved visual experience, particularly when working with both volumetric images and 3D objects. Our method has been successfully implemented with one virtual reality (VR) headset and one augmented reality (AR) headset with satisfying results achieved. We also conducted two user studies and proved the effectiveness of the interactors and the efficiency of our method in comparison with other approaches for neuron tracing.

Background: Material characterization of brain white matter (BWM) is difficult due to the anisotropy inherent to the three-dimensional microstructure and the various interactions between heterogeneous brain-tissue (axon, myelin, and glia). Developing full scale finite element models that accurately represent the relationship between the micro and macroscale BWM is however extremely challenging and computationally expensive. The anisotropic properties of the microstructure of BWM computed by building unit cells under frequency domain viscoelasticity comprises of 36 individual constants each, for the loss and storage moduli. Furthermore, the architecture of each unit cell is arbitrary in an infinite dataset., Methods: In this study, we extend our previous work on developing representative volume elements (RVE) of the microstructure of the BWM in the frequency domain to develop 3D deep learning algorithms that can predict the anisotropic composite properties. The deep 3D convolutional neural network (CNN) algorithms utilizes a voxelization method to obtain geometry information from 3D RVEs. The architecture information encoded in the voxelized location is employed as input data while cross-referencing the RVEs' material properties (output data). We further develop methods by incorporating parallel pathways, Residual Neural Networks and inception modulus that improve the efficiency of deep learning algorithms., Results: This paper presents different CNN algorithms in predicting the anisotropic composite properties of BWM. A quantitative analysis of the individual algorithms is presented with the view of identifying optimal strategies to interpret the combined measurements of brain MRE and DTI., Significance: The proposed Multiscale 3D ResNet (M3DR) algorithm demonstrates high learning ability and performance over baseline CNN algorithms in predicting BWM tissue properties. The hybrid M3DR framework also overcomes the significant limitations encountered in modeling brain tissue using finite elements alone including those such as high computational cost, mesh and simulation failure. The proposed framework also provides an efficient and streamlined platform for implementing complex boundary conditions, modeling intrinsic material properties and imparting interfacial architecture information., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 The Authors. Published by Elsevier B.V. All rights reserved.)

Background and Objective: The incidence of facial fractures is on the rise globally, yet limited studies are addressing the diverse forms of facial fractures present in 3D images. In particular, due to the nature of the facial fracture, the direction in which the bone fractures vary, and there is no clear outline, it is difficult to determine the exact location of the fracture in 2D images. Thus, 3D image analysis is required to find the exact fracture area, but it needs heavy computational complexity and expensive pixel-wise labeling for supervised learning. In this study, we tackle the problem of reducing the computational burden and increasing the accuracy of fracture localization by using a weakly-supervised object localization without pixel-wise labeling in a 3D image space., Methods: We propose a Very Fast, High-Resolution Aggregation 3D Detection CAM (VFHA-CAM) model, which can detect various facial fractures. To better detect tiny fractures, our model uses high-resolution feature maps and employs Ablation CAM to find an exact fracture location without pixel-wise labeling, where we use a rough fracture image detected with 3D box-wise labeling. To this end, we extract important features and use only essential features to reduce the computational complexity in 3D image space., Results: Experimental findings demonstrate that VFHA-CAM surpasses state-of-the-art 2D detection methods by up to 20% in sensitivity/person and specificity/person, achieving sensitivity/person and specificity/person scores of 87% and 85%, respectively. In addition, Our VFHA-CAM reduces location analysis time to 76 s without performance degradation compared to a simple Ablation CAM method that takes more than 20 min., Conclusion: This study introduces a novel weakly-supervised object localization approach for bone fracture detection in 3D facial images. The proposed method employs a 3D detection model, which helps detect various forms of facial bone fractures accurately. The CAM algorithm adopted for fracture area segmentation within a 3D fracture detection box is key in quickly informing medical staff of the exact location of a facial bone fracture in a weakly-supervised object localization. In addition, we provide 3D visualization so that even non-experts unfamiliar with 3D CT images can identify the fracture status and location., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 ZIOVISION co. ltd. Published by Elsevier B.V. All rights reserved.)

Purpose: Non-contrast-enhanced time of flight (TOF) is a standard method for magnetic resonance angiography used to depict vessel morphology. TOF is commonly performed with a 3D steady-state acquisition, employing a short repetition time to support high resolution imaging. At 7 T, TOF exhibits substantial increase in SNR and contrast, improving its clinical value. However, one of the remaining challenges, exacerbated at 7 T, is the presence of artifacts due to pulsatile blood flow, especially near major blood vessels. In this study we examine a method to significantly reduce these artifacts., Methods: We recently introduced a new "local-scrambling" approach that semi-randomizes the acquisition order of the phase encodes, to achieve a controllable cutoff frequency above which the artifacts are drastically reduced. With this approach, artifacts resulting from fast local fluctuations such as cardiac pulsation are significantly reduced. In this study, we explore the ability of this local-scrambling approach to reduce pulsatile blood flow artifacts in a 3D TOF acquisition. Cartesian line-by-line and center-out ordering, with and without local-scrambling, were compared in simulations and in human brain imaging at 3 and 7 T scanners., Results: In the simulations the artifact intensity showed a 10-fold reduction using local-scrambling compared to line-by-line and 4-fold compared to center-out ordering. In vivo results show that artifacts are much more pronounced at 7 T compared to 3 T, and in both cases they are effectively reduced by local-scrambling., Conclusion: Local-scrambling improves image quality for both line-by-line and center-out ordering. This approach can easily be implemented in the scanner without any changes to the reconstruction., (© 2024 The Author(s). Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.)

Purpose: To reduce the ringing artifacts of the motion-resolved images in free-breathing dynamic pulmonary MRI., Methods: A golden-step based interleaving (GSI) technique was proposed to reduce ringing artifacts induced by diaphragm drifting. The pulmonary MRI data were acquired using a superior-inferior navigated 3D radial UTE sequence in an interleaved manner during free breathing. Successive interleaves were acquired in an incoherent fashion along the polar direction. Four-dimensional images were reconstructed from the motion-resolved k-space data obtained by retrospectively binning. The reconstruction algorithms included standard nonuniform fast Fourier transform (NUFFT), Voronoi-density-compensated NUFFT, extra-dimensional UTE, and motion-state weighted motion-compensation reconstruction. The proposed interleaving technique was compared with a conventional sequential interleaving (SeqI) technique on a phantom and eight subjects., Results: The quantified ringing artifacts level in the motion-resolved image is positively correlated with the quantified nonuniformity level of the corresponding k-space. The nonuniformity levels of the end-expiratory and end-inspiratory k-space binned from GSI data (0.34 ± 0.07, 0.33 ± 0.05) are significantly lower with statistical significance (p < 0.05) than that binned from SeqI data (0.44 ± 0.11, 0.42 ± 0.12). Ringing artifacts are substantially reduced in the dynamic images of eight subjects acquired using the proposed technique in comparison with that acquired using the conventional SeqI technique., Conclusion: Ringing artifacts in the motion-resolved images induced by diaphragm drifting can be reduced using the proposed GSI technique for free-breathing dynamic pulmonary MRI. This technique has the potential to reduce ringing artifacts in free-breathing liver and kidney MRI based on full-echo interleaved 3D radial acquisition., (© 2024 International Society for Magnetic Resonance in Medicine.)

Purpose: To introduce an alternative idea for fat suppression that is suited both for low-field applications where conventional fat-suppression approaches become ineffective due to narrow spectral separation and for applications with strong B0 homogeneities., Methods: Separation of fat and water is achieved by sweeping the frequency of RF saturation pulses during continuous radial acquisition and calculating frequency-resolved images using regularized iterative reconstruction. Voxel-wise signal-response curves are extracted that reflect tissue's response to RF saturation at different frequencies and allow the classification into fat or water. This information is then utilized to generate water-only composite images. The principle is demonstrated in free-breathing abdominal and neck examinations using stack-of-stars 3D balanced SSFP (bSSFP) and gradient-recalled echo (GRE) sequences at 0.55 and 3T. Moreover, a potential extension toward quantitative fat/water separation is described., Results: Experiments with a proton density fat fraction (PDFF) phantom validated the reliability of fat/water separation using signal-response curves. As demonstrated for abdominal imaging at 0.55T, the approach resulted in more uniform fat suppression without loss of water signal and in improved CSF-to-fat signal ratio. Moreover, the approach provided consistent fat suppression in 3T neck exams where conventional spectrally-selective fat saturation failed due to strong local B0 inhomogeneities. The feasibility of simultaneous fat/water quantification has been demonstrated in a PDFF phantom., Conclusion: The proposed principle achieves reliable fat suppression in low-field applications and adapts to high-field applications with strong B0 inhomogeneity. Moreover, the principle potentially provides a basis for developing an alternative approach for PDFF quantification., (© 2024 International Society for Magnetic Resonance in Medicine.)

Purpose: This research aimed to develop an innovative method for designing and fabricating nasal prostheses that reduces anaplastologist expertise dependency while maintaining quality and appearance, allowing patients to regain their normal facial appearance., Methods: The method involved statistical shape modeling using a morphable face model and 3D data acquired through optical scanning or CT. An automated design process generated patient-specific fits and appearances using regular prosthesis materials and 3D printing of molds. Manual input was required for specific case-related details., Results: The developed method met all predefined requirements, replacing analog impression-making and offering compatibility with various data acquisition methods. Prostheses created through this method exhibited equivalent aesthetics to conventionally fabricated ones while reducing the skill dependency typically associated with prosthetic design and fabrication., Conclusions: This method provides a promising approach for both temporary and definitive nasal prostheses, with the potential for remote prosthesis fabrication in areas lacking anaplastology care. While new skills are required for data acquisition and algorithm control, these technologies are increasingly accessible. Further clinical studies will help validate its effectiveness, and ongoing technological advancements may lead to even more advanced and skill-independent prosthesis fabrication methods in the future., (© 2024. The Author(s).)

Rationale and Objectives: Cardiac magnetic resonance imaging is a crucial tool for analyzing, diagnosing, and formulating treatment plans for cardiovascular diseases. Currently, there is very little research focused on balancing cardiac segmentation performance with lightweight methods. Despite the existence of numerous efficient image segmentation algorithms, they primarily rely on complex and computationally intensive network models, making it challenging to implement them on resource-constrained medical devices. Furthermore, simplified models designed to meet the requirements of device lightweighting may have limitations in comprehending and utilizing both global and local information for cardiac segmentation., Materials and Methods: We propose a novel 3D high-performance lightweight medical image segmentation network, HL-UNet, for application in cardiac image segmentation. Specifically, in HL-UNet, we propose a novel residual-enhanced Adaptive attention (REAA) module that combines residual-enhanced connectivity with an adaptive attention mechanism to efficiently capture key features of input images and optimize their representation capabilities, and integrates the Visual Mamba (VSS) module to enhance the performance of HL-UNet., Results: Compared to large-scale models such as TransUNet, HL-UNet increased the Dice of the right ventricular cavity (RV), left ventricular myocardia (MYO), and left ventricular cavity (LV), the key indicators of cardiac image segmentation, by 1.61%, 5.03% and 0.19%, respectively. At the same time, the Params and FLOPs of the model decreased by 41.3 M and 31.05 G, respectively. Furthermore, compared to lightweight models such as the MISSFormer, the HL-UNet improves the Dice of RV, MYO, and LV by 4.11%, 3.82%, and 4.33%, respectively, when the number of parameters and computational complexity are close to or even lower., Conclusion: The proposed HL-UNet model captures local details and edge information in images while being lightweight. Experimental results show that compared with large-scale models, HL-UNet significantly reduces the number of parameters and computational complexity while maintaining performance, thereby increasing frames per second (FPS). Compared to lightweight models, HL-UNet shows substantial improvements across various key metrics, with parameter count and computational complexity approaching or even lower., Competing Interests: Declaration of Competing Interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Daji Ergu reports financial support was provided by National Natural Science Foundation of China. Hai Zhou reports financial support was provided by Southwest Minzu University. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.)

Purpose: To develop and demonstrate a fast 3D fMRI acquisition technique with high spatial resolution over a reduced FOV, named k-t 3D reduced FOV imaging (3D-rFOVI)., Methods: Based on 3D gradient-echo EPI, k-t 3D-rFOVI used a 2D RF pulse to reduce the FOV in the in-plane phase-encoding direction, boosting spatial resolution without increasing echo train length. For image acceleration, full sampling was applied in the central k-space region along the through-slab direction (k z ) for all time frames, while randomized undersampling was used in outer k z regions at different time frames. Images were acquired at 3T and reconstructed using a method based on partial separability. fMRI detection sensitivity of k-t 3D-rFOVI was quantitively analyzed with simulation data. Human visual fMRI experiments were performed to evaluate k-t 3D-rFOVI and compare it with a commercial multiband EPI sequence., Results: The simulation data showed that k-t 3D-rFOVI can detect 100% of fMRI activations with an acceleration factor (R) of 2 and ˜80% with R = 6. In the human fMRI data acquired with 1.5-mm spatial resolution and 800-ms volume TR (TR vol ), k-t 3D-rFOVI with R = 4 detected 46% more activated voxels in the visual cortex than the multiband EPI. Additional fMRI experiments showed that k-t 3D-rFOVI can achieve TR vol of 480 ms with R = 6, while reliably detecting visual activation., Conclusions: k-t 3D-rFOVI can simultaneously achieve a high spatial resolution (1.5-mm isotropically) and short TR vol (480-ms) at 3T. It offers a robust acquisition technique for fast fMRI studies over a focused brain volume., (© 2024 The Author(s). Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.)

Experimental blood flow measurement techniques are invaluable for a better understanding of cardiovascular disease formation, progression, and treatment. One of the emerging methods is time-resolved three-dimensional phase-contrast magnetic resonance imaging (4D flow MRI), which enables noninvasive time-dependent velocity measurements within large vessels. However, several limitations hinder the usability of 4D flow MRI and other experimental methods for quantitative hemodynamics analysis. These mainly include measurement noise, corrupt or missing data, low spatiotemporal resolution, and other artifacts. Traditional filtering is routinely applied for denoising experimental blood flow data without any detailed discussion on why it is preferred over other methods. In this study, filtering is compared to different singular value decomposition (SVD)-based machine learning and autoencoder-type deep learning methods for denoising and filling in missing data (imputation). An artificially corrupted and voxelized computational fluid dynamics (CFD) simulation as well as in vitro 4D flow MRI data are used to test the methods. SVD-based algorithms achieve excellent results for the idealized case but severely struggle when applied to in vitro data. The autoencoders are shown to be versatile and applicable to all investigated cases. For denoising, the in vitro 4D flow MRI data, the denoising autoencoder (DAE), and the Noise2Noise (N2N) autoencoder produced better reconstructions than filtering both qualitatively and quantitatively. Deep learning methods such as N2N can result in noise-free velocity fields even though they did not use clean data during training. This work presents one of the first comprehensive assessments and comparisons of various classical and modern machine-learning methods for enhancing corrupt cardiovascular flow data in diseased arteries for both synthetic and experimental test cases., (© 2024 The Author(s). International Journal for Numerical Methods in Biomedical Engineering published by John Wiley & Sons Ltd.)

Stylization of 3D shapes has been a prominent topic in computer graphics. Geometric stylization, as a category within 3D shape stylization, has garnered widespread attention due to its distinctive geometric aesthetics. This paper introduces a novel and simplified method for achieving regular octahedral stylization, capable of transforming the input shape into the regular octahedral style while preserving the content of the original shape. The proposed method achieves regular octahedral stylization by using face normals. Its implementation is straightforward because it incurs a cost only in terms of solving several linear systems. To demonstrate the effectiveness and robustness of the proposed method, extensive tests were conducted on models with varying levels of complexity and topological structures. The results highlight the exceptional capability of the proposed method to generate geometric shapes with an aesthetically pleasing regular octahedral style., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2024 Dong et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

The reflectance confocal microscopy (RCM)-optical coherence tomography (OCT) device has shown utility in detecting and assessing the depth of basal cell carcinoma (BCC) in vivo but is challenging for novices to interpret. Artificial intelligence (AI) applied to RCM-OCT could aid readers. We trained AI models, using OCT rasters of biopsy-confirmed BCC, to detect BCC, create three-dimensional rendering and automatically measure tumour depth. Trained AI models were applied to a separate test set containing rasters of BCC, benign lesions, and healthy skin. Blinded reader analysis and tumour depth correlation with histopathology were conducted. BCC detection improved from viewing OCT rasters only (sensitivity 73.3%, specificity 45.5%) to viewing rasters with AI-generated BCC rendering (sensitivity 86.7%, specificity 48.5%). A Pearson correlation r2 = 0.59 (P = 0.02) was achieved for the tumour depth measurement between AI and histological measured depths. Thus, addition of AI to the RCM-OCT device may expand its utility widely., Competing Interests: Conflicts of interest The authors declare no conflicts of interest., (© The Author(s) 2024. Published by Oxford University Press on behalf of British Association of Dermatologists. All rights reserved. For commercial re-use, please contact reprints@oup.com for reprints and translation rights for reprints. All other permissions can be obtained through our RightsLink service via the Permissions link on the article page on our site—for further information please contact journals.permissions@oup.com.)

Cinematic Rendering (CR) employs physical models such as ray tracing and global illumination to simulate real-world light phenomena, producing high-quality images with rich details. In the medical field, CR can significantly aid doctors in accurate diagnosis and preoperative planning. However, doctors require efficient real-time rendering when using CR, which presents a challenge due to the substantial computing resources demanded by CR's ray tracing and global illumination models. Precomputed lighting can enhance the efficiency of real-time rendering by freezing certain scene variables. Typically, precomputed methods freeze geometry and materials. However, since the physical rendering of medical images relies on volume data rendering of transfer functions, the CR algorithm cannot utilize precomputed methods directly. To improve the rendering efficiency of the CR algorithm, we propose a precomputed low-frequency lighting method. By simulating the lighting pattern of shadowless surgical lamps, we adopt a spherical distribution of multiple light sources, with each source capable of illuminating the entire volume of data. Under the influence of these large-area multi-light sources, the precomputed lighting adheres to physical principles, resulting in shadow-free and uniformly distributed illumination. We integrated this precomputed method into the ray-casting algorithm, creating an accelerated CR algorithm that achieves more than twice the rendering efficiency of traditional CR rendering., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2024 Yuan et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

As an alternative to true isotropic 3D imaging, image super-resolution (SR) has been applied to reconstruct an isotropic 3D volume from multiple anisotropic scans. However, traditional SR methods struggle with inadequate performance, prolonged processing times, and the necessity for manual feature extraction. Motivated by the exceptional representational ability and automatic feature extraction of convolutional neural networks (CNNs), in this work, we present an end-to-end isotropic MRI reconstruction strategy based on deep learning. The proposed method is based on 3D convolutional neural networks (3D CNNs), which can effectively capture the 3D structural features of MRI volumes and accurately predict potential structure. In addition, the proposed method takes multiple orthogonal scans as input and thus enables the model to use more complementary information from different dimensions for precise inference. Experimental results show that the proposed algorithm achieves promising performance in terms of both quantitative and qualitative assessments. In addition, it can process a 3D volume with a size of 256 × 256 × 256 in less than 1 min with the support of an NVIDIA GeForce GTX 1080Ti GPU, which suggests that it is not only a quantitatively superior method but also a practical one.

The estimation of vessels' centerlines is a critical step in assessing the geometry of the vessel, the topological representation of the vessel tree, and vascular network visualization. In this research, we present a novel method for obtaining geometric parameters from peripheral arteries in 3D medical binary volumes. Our approach focuses on centerline extraction, which yields smooth and robust results. The procedure starts with a segmented 3D binary volume, from which a distance map is generated using the Euclidean distance transform. Subsequently, a skeleton is extracted, and seed points and endpoints are identified. A search methodology is used to derive the best path on the skeletonized 3D binary array while tracking from the goal points to the seed point. We use the distance transform to calculate the distance between voxels and the nearest vessel surface, while also addressing bifurcations when vessels divide into multiple branches. The proposed method was evaluated on 22 real cases and 10 synthetically generated vessels. We compared our method to different state-of-the-art approaches and demonstrated its better performance. The proposed method achieved an average error of 1.382 mm with real patient data and 0.571 mm with synthetic data, both of which are lower than the errors obtained by other state-of-the-art methodologies. This extraction of the centerline facilitates the estimation of multiple geometric parameters of vessels, including radius, curvature, and length.

Object: In a typical MR session, several contrasts are acquired. Due to the sequential nature of the data acquisition process, the patient may experience some discomfort at some point during the session, and start moving. Hence, it is quite common to have MR sessions where some contrasts are well-resolved, while other contrasts exhibit motion artifacts. Instead of repeating the scans that are corrupted by motion, we introduce a reference-guided retrospective motion correction scheme that takes advantage of the motion-free scans, based on a generalized rigid registration routine., Materials and Methods: We focus on various existing clinical 3D brain protocols at 1.5 Tesla MRI based on Cartesian sampling. Controlled experiments with three healthy volunteers and three levels of motion are performed., Results: Radiological inspection confirms that the proposed method consistently ameliorates the corrupted scans. Furthermore, for the set of specific motion tests performed in this study, the quality indexes based on PSNR and SSIM shows only a modest decrease in correction quality as a function of motion complexity., Discussion: While the results on controlled experiments are positive, future applications to patient data will ultimately clarify whether the proposed correction scheme satisfies the radiological requirements., (© 2024. The Author(s).)

Purpose: To achieve high-resolution fetal brain anatomical imaging without introducing image artifacts by reducing the FOV, and to demonstrate improved image quality compared to conventional full-FOV fetal brain imaging., Methods: Reduced FOV was achieved by applying outer volume suppression (OVS) pulses immediately prior to standard single-shot fast spin echo (SSFSE) imaging. In the OVS preparation, a saturation RF pulse followed by a gradient spoiler was repeated three times with optimized flip-angle weightings and a variable spoiler scheme to enhance signal suppression. Simulations and phantom and in-vivo experiments were performed to evaluate OVS performance. In-vivo high-resolution SSFSE images acquired using the proposed approach were compared with conventional and high-resolution SSFSE images with a full FOV, using image quality scores assessed by neuroradiologists and calculated image metrics., Results: Excellent signal suppression in the saturation bands was confirmed in phantom and in-vivo experiments. High-resolution SSFSE images with a reduced FOV acquired using OVS demonstrated the improved depiction of brain structures without significant motion and blurring artifacts. The proposed method showed the highest image quality scores in the criteria of sharpness, contrast, and artifact and was selected as the best method based on overall image quality. The calculated image sharpness and tissue contrast ratio were also the highest with the proposed method., Conclusion: High-resolution fetal brain anatomical images acquired using a reduced FOV with OVS demonstrated improved image quality both qualitatively and quantitatively, suggesting the potential for enhanced diagnostic accuracy in detecting fetal brain abnormalities in utero., (© 2024 International Society for Magnetic Resonance in Medicine.)

The self-supervised monocular depth estimation framework is well-suited for medical images that lack ground-truth depth, such as those from digestive endoscopes, facilitating navigation and 3D reconstruction in the gastrointestinal tract. However, this framework faces several limitations, including poor performance in low-texture environments, limited generalisation to real-world datasets, and unclear applicability in downstream tasks like visual servoing. To tackle these challenges, we propose MonoLoT, a self-supervised monocular depth estimation framework featuring two key innovations: point matching loss and batch image shuffle. Extensive ablation studies on two publicly available datasets, namely C3VD and SimCol, have shown that methods enabled by MonoLoT achieve substantial improvements, with accuracies of 0.944 on C3VD and 0.959 on SimCol, surpassing both depth-supervised and self-supervised baselines on C3VD. Qualitative evaluations on real-world endoscopic data underscore the generalisation capabilities of our methods, outperforming both depth-supervised and self-supervised baselines. To demonstrate the feasibility of using monocular depth estimation for visual servoing, we have successfully integrated our method into a proof-of-concept robotic platform, enabling real-time automatic intervention and control in digestive endoscopy. In summary, our method represents a significant advancement in monocular depth estimation for digestive endoscopy, overcoming key challenges and opening promising avenues for medical applications.

Purpose: This study aimed to facilitate research progress in MR elastography (MRE) by providing a versatile and convenient application for MRE reconstruction, namely the MRE research tool (MRE-rTool). It can be used for a series of MRE image analyses, including phase unwrapping, arbitrary bandpass and directional filtering, noise assessment of the wave propagation image (motion SNR), and reconstruction of the elastogram in both 2D and 3D MRE acquisitions. To reinforce the versatility of MRE-rTool, the conventional method of motion SNR was modified into a new method that reflects the effects of image filtering., Methods: MRE tests of the phantom and liver were performed using different estimation algorithms for stiffness value (algebraic inversion of the differential equation [AIDE], local frequency estimation [LFE] in MRE-rTool, and multimodel direct inversion [MMDI] in clinical reconstruction) and acquiring dimensions (2D and 3D acquisitions). This study also tested the accuracy of masking low SNR regions using modified and conventional motion SNR under various mechanical vibration powers., Results: The stiffness values estimated using AIDE/LFE in MRE-rTool were comparable to that of MMDI (phantom, 3.71 ± 0.74, 3.60 ± 0.32, and 3.60 ± 0.54 kPa in AIDE, LFE, and MMDI; liver, 2.26 ± 0.31, 2.74 ± 0.16, and 2.21 ± 0.26 kPa in AIDE, LFE, and MMDI). The stiffness value in 3D acquisition was independent of the direction of the motion-encoding gradient and was more accurate than that of 2D acquisition. The masking of low SNR regions using the modified motion SNR worked better than that in the conventional motion SNR for each vibration power, especially when using a directional filter., Conclusion: The performance of MRE-rTool on test data reached the level required in clinical MRE studies. MRE-rTool has the potential to facilitate MRE research, contribute to the future development of MRE, and has been freely released online.

We introduce the largest abdominal CT dataset (termed AbdomenAtlas) of 20,460 three-dimensional CT volumes sourced from 112 hospitals across diverse populations, geographies, and facilities. AbdomenAtlas provides 673 K high-quality masks of anatomical structures in the abdominal region annotated by a team of 10 radiologists with the help of AI algorithms. We start by having expert radiologists manually annotate 22 anatomical structures in 5,246 CT volumes. Following this, a semi-automatic annotation procedure is performed for the remaining CT volumes, where radiologists revise the annotations predicted by AI, and in turn, AI improves its predictions by learning from revised annotations. Such a large-scale, detailed-annotated, and multi-center dataset is needed for two reasons. Firstly, AbdomenAtlas provides important resources for AI development at scale, branded as large pre-trained models, which can alleviate the annotation workload of expert radiologists to transfer to broader clinical applications. Secondly, AbdomenAtlas establishes a large-scale benchmark for evaluating AI algorithms-the more data we use to test the algorithms, the better we can guarantee reliable performance in complex clinical scenarios. An ISBI & MICCAI challenge named BodyMaps: Towards 3D Atlas of Human Body was launched using a subset of our AbdomenAtlas, aiming to stimulate AI innovation and to benchmark segmentation accuracy, inference efficiency, and domain generalizability. We hope our AbdomenAtlas can set the stage for larger-scale clinical trials and offer exceptional opportunities to practitioners in the medical imaging community. Codes, models, and datasets are available at https://www.zongweiz.com/dataset., Competing Interests: Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Zongwei Zhou reports financial support was provided by Johns Hopkins University. Zongwei Zhou reports a relationship with Johns Hopkins University that includes: employment and funding grants. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier B.V. All rights reserved.)

Purpose: To develop a generalized rigid body motion correction method in 3D radial brain MRI to deal with continuous motion pattern through projection moment analysis., Methods: An assumption was made that the multichannel coil moves with the head, which was achieved by using a flexible head coil. A two-step motion correction scheme was proposed to directly extract the motion parameters from the acquired k-space data using the analysis of center-of-mass with high noise robustness, which were used for retrospective motion correction. A recursive least-squares model was introduced to recursively estimate the motion parameters for every single spoke, which used the smoothness of motion and resulted in high temporal resolution and low computational cost. Five volunteers were scanned at 3 T using a 3D radial multidimensional golden-means trajectory with instructed motion patterns. The performance was tested through both simulation and in vivo experiments. Quantitative image quality metrics were calculated for comparison., Results: The proposed method showed good accuracy and precision in both translation and rotation estimation. A better result was achieved using the proposed two-step correction compared to traditional one-step correction without significantly increasing computation time. Retrospective correction showed substantial improvements in image quality among all scans, even for stationary scans., Conclusions: The proposed method provides an easy, robust, and time-efficient tool for motion correction in brain MRI, which may benefit clinical diagnosis of uncooperative patients as well as scientific MRI researches., (© 2024 International Society for Magnetic Resonance in Medicine.)

Purpose: To develop a practical method to enable 3D T 1 mapping of brain metabolites., Theory and Methods: Due to the high dimensionality of the imaging problem underlying metabolite T 1 mapping, measurement of metabolite T 1 values has been currently limited to a single voxel or slice. This work achieved 3D metabolite T 1 mapping by leveraging a recent ultrafast MRSI technique called SPICE (spectroscopic imaging by exploiting spatiospectral correlation). The Ernst-angle FID MRSI data acquisition used in SPICE was extended to variable flip angles, with variable-density sparse sampling for efficient encoding of metabolite T 1 information. In data processing, a novel generalized series model was used to remove water and subcutaneous lipid signals; a low-rank tensor model with prelearned subspaces was used to reconstruct the variable-flip-angle metabolite signals jointly from the noisy data., Results: The proposed method was evaluated using both phantom and healthy subject data. Phantom experimental results demonstrated that high-quality 3D metabolite T 1 maps could be obtained and used for correction of T 1 saturation effects. In vivo experimental results showed metabolite T 1 maps with a large spatial coverage of 240 × 240 × 72 mm 3 and good reproducibility coefficients (< 11%) in a 14.5-min scan. The metabolite T 1 times obtained ranged from 0.99 to 1.44 s in gray matter and from 1.00 to 1.35 s in white matter., Conclusion: We successfully demonstrated the feasibility of 3D metabolite T 1 mapping within a clinically acceptable scan time. The proposed method may prove useful for both T 1 mapping of brain metabolites and correcting the T 1 -weighting effects in quantitative metabolic imaging., (© 2024 The Author(s). Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.)

Purpose: To develop a new sequence to simultaneously acquire Cartesian sodium ( 23 Na) MRI and accelerated Cartesian single (SQ) and triple quantum (TQ) sodium MRI of in vivo human brain at 7 T by leveraging two dedicated low-rank reconstruction frameworks., Theory and Methods: The Double Half-Echo technique enables short echo time Cartesian 23 Na MRI and acquires two k-space halves, reconstructed by a low-rank coupling constraint. Additionally, three-dimensional (3D) 23 Na Multi-Quantum Coherences (MQC) MRI requires multi-echo sampling paired with phase-cycling, exhibiting a redundant multidimensional space. Simultaneous Autocalibrating and k-Space Estimation (SAKE) were used to reconstruct highly undersampled 23 Na MQC MRI. Reconstruction performance was assessed against five-dimensional (5D) CS, evaluating structural similarity index (SSIM), root mean squared error (RMSE), signal-to-noise ratio (SNR), and quantification of tissue sodium concentration and TQ/SQ ratio in silico, in vitro, and in vivo., Results: The proposed sequence enabled the simultaneous acquisition of fully sampled 23 Na MRI while leveraging prospective undersampling for 23 Na MQC MRI. SAKE improved TQ image reconstruction regarding SSIM by 6% and reduced RMSE by 35% compared to 5D CS in vivo. Thanks to prospective undersampling, the spatial resolution of 23 Na MQC MRI was enhanced from 8 × 8 × 15 $$ 8\times 8\times 15 $$ mm 3 to 8 × 8 × 8 $$ 8\times 8\times 8 $$ mm 3 while reducing acquisition time from 2 × 31 $$ 2\times 31 $$ min to 2 × 23 $$ 2\times 23 $$ min., Conclusion: The proposed sequence, coupled with low-rank reconstructions, provides an efficient framework for comprehensive whole-brain sodium MRI, combining TSC, T2*, and TQ/SQ ratio estimations. Additionally, low-rank matrix completion enables the reconstruction of highly undersampled 23 Na MQC MRI, allowing for accelerated acquisition or enhanced spatial resolution., (© 2024 The Authors. Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.)

Introduction: Research regarding orthodontic changes using the superimposition of digital study models (DSMs) is commonplace. Information regarding the accuracy of data processing by superimposition software is limited. The study aimed to compare different methods of superimposing DSMs using implant-supported crowns (ISC) as a stable reference structure., Methods: DSMs containing ISCs were sourced from a database of patients treated with clear aligner therapy. The DSM representing the planned treatment outcome was superimposed on the pretreatment DSM. Three tooth points were selected for comparison on the contralateral side of each ISC. Differences in Cartesian coordinates for each tooth point for each arch superimposition method, used by the Geomagic Control X (3D systems, Rock Hill, NC) software system, were recorded. Paired t tests for the reference standard superimposition method best-fit high-resolution using the entire dental arch compared with initial, best-fit low-resolution, and best-fit high-resolution using the ISC only were calculated., Results: The DSMs of 54 dental arches containing ISCs were evaluated. All mean differences for displacements of selected points on the contralateral side to the ISC in the 3 Cartesian planes were <0.05 mm (P <0.05) and below the threshold of clinical significance. In addition, the standard superimposition techniques (initial, best-fit low-resolution, and best-fit high-resolution) resulted in nonstatistically significant and nonclinically significant differences in the position of the ISC., Conclusions: Researchers can be confident that the described superimposition methodologies, with and without ISCs as a stable reference structure, are a valid method for accurately assessing most intraarch dental changes., (Copyright © 2024 American Association of Orthodontists. Published by Elsevier Inc. All rights reserved.)

This paper introduces an innovative methodology for producing high-quality 3D lung CT images guided by textual information. While diffusion-based generative models are increasingly used in medical imaging, current state-of-the-art approaches are limited to low-resolution outputs and underutilize radiology reports' abundant information. The radiology reports can enhance the generation process by providing additional guidance and offering fine-grained control over the synthesis of images. Nevertheless, expanding text-guided generation to high-resolution 3D images poses significant memory and anatomical detail-preserving challenges. Addressing the memory issue, we introduce a hierarchical scheme that uses a modified UNet architecture. We start by synthesizing low-resolution images conditioned on the text, serving as a foundation for subsequent generators for complete volumetric data. To ensure the anatomical plausibility of the generated samples, we provide further guidance by generating vascular, airway, and lobular segmentation masks in conjunction with the CT images. The model demonstrates the capability to use textual input and segmentation tasks to generate synthesized images. Algorithmic comparative assessments and blind evaluations conducted by 10 board-certified radiologists indicate that our approach exhibits superior performance compared to the most advanced models based on GAN and diffusion techniques, especially in accurately retaining crucial anatomical features such as fissure lines and airways. This innovation introduces novel possibilities. This study focuses on two main objectives: (1) the development of a method for creating images based on textual prompts and anatomical components, and (2) the capability to generate new images conditioning on anatomical elements. The advancements in image generation can be applied to enhance numerous downstream tasks.

Volume reconstruction techniques are gaining increasing interest in medical domains due to their potential to learn complex 3D structural information from sparse 2D images. Recently, neural radiance fields (NeRF), which implicitly model continuous radiance fields based on multi-layer perceptrons to enable volume reconstruction of objects at arbitrary resolution, have gained traction in natural image volume reconstruction. However, the direct application of NeRF to medical volume reconstruction presents unique challenges due to differences in imaging principles, internal structure requirements, and boundary delineation. In this paper, we evaluate different NeRF techniques developed for natural images, including sampling strategies, feature encoding, and the use of complimentary features, by applying them to medical images. We evaluate three state-of-the-art NeRF techniques on four datasets of medical images of different complexity. Our goal is to identify the strengths, limitations, and future directions for integrating NeRF into the medical domain.

Purpose: The conventional method to reconstruct the bone level for orbital defects, which is based on mirroring and manual adaptation, is time-consuming and the accuracy highly depends on the expertise of the clinical engineer. The aim of this study is to propose and evaluate an automated reconstruction method utilizing a Gaussian process morphable model (GPMM)., Methods: Sixty-five Computed Tomography (CT) scans of healthy midfaces were used to create a GPMM that can model shape variations of the orbital region. Parameter optimization was performed by evaluating several quantitative metrics inspired on the shape modeling literature, e.g. generalization and specificity. The reconstruction error was estimated by reconstructing artificial defects created in orbits from fifteen CT scans that were not included in the GPMM. The developed algorithms utilize the existing framework of Gaussian process morphable models, as implemented in the Scalismo software., Results: By evaluating the proposed quality metrics, adequate parameters are chosen for non-rigid registration and reconstruction. The resulting median reconstruction error using the GPMM was lower (0.35 ± 0.16 mm) compared to the mirroring method (0.52 ± 0.18 mm). In addition, the GPMM-based reconstruction is automated and can be applied to large bilateral defects with a median reconstruction error of 0.39 ± 0.11 mm., Conclusion: The GPMM-based reconstruction proves to be less time-consuming and more accurate than reconstruction by mirroring. Further validation through clinical studies on patients with orbital defects is warranted. Nevertheless, the results underscore the potential of GPMM-based reconstruction as a promising alternative for designing patient-specific implants., (© 2024. CARS.)

Purpose: To develop acceleration strategies for 3D multi-slab diffusion weighted imaging (3D ms-DWI) for enabling applications that require simultaneously high spatial (1 mm isotropic) and angular (> 30 directions) resolutions., Methods: 3D ms-DWI offers high SNR-efficiency, with the ability to achieve high isotropic spatial resolution, yet suffers from long scan-times for studies requiring high angular resolutions. We develop 6D k-q space acceleration strategies to reduce the scan-time. Specifically, we develop non-uniform 3D k y -k z under-sampling employing a shot-selective 2D CAIPI sampling approach. To achieve inter-shot phase-compensation, 2D navigators were employed that utilize the same CAIPI trajectory. An iterative model-based 3D multi-shot reconstruction was designed by incorporating phase into the forward encoding process. Additionally, the shot-selective non-uniform k y -k z CAIPI acceleration was randomized along the q-dimension. The 3D model-based multi-shot reconstruction is then extended to a joint reconstruction that simultaneously reconstructs all the q-space points, with the help of a spatial total variation and deep-learned q-space regularization., Results: The proposed reconstruction is shown to achieve adequate phase-compensation in both 2D CAIPI accelerated and additional k y -k z under-sampled cases. Using retrospective under-sampling experiments, we show that k-q accelerations close a factor of 12 can be achieved with a reconstruction error < 3% for both single and multi-shell data. This translates to a scan-time reduction by 3-fold for experiments with simultaneously high spatial and angular resolutions., Conclusion: The proposed method facilitates the utilization of 3D ms-DWI for simultaneously high k-q resolution applications with close to 3× reduced scan-time., (Copyright © 2024 Elsevier Inc. All rights reserved.)

Purpose: To present and assess an outlier mitigation method that makes free-running volumetric cardiovascular MRI (CMR) more robust to motion., Methods: The proposed method, called compressive recovery with outlier rejection (CORe), models outliers in the measured data as an additive auxiliary variable. We enforce MR physics-guided group sparsity on the auxiliary variable, and jointly estimate it along with the image using an iterative algorithm. For evaluation, CORe is first compared to traditional compressed sensing (CS), robust regression (RR), and an existing outlier rejection method using two simulation studies. Then, CORe is compared to CS using seven three-dimensional (3D) cine, 12 rest four-dimensional (4D) flow, and eight stress 4D flow imaging datasets., Results: Our simulation studies show that CORe outperforms CS, RR, and the existing outlier rejection method in terms of normalized mean square error and structural similarity index across 55 different realizations. The expert reader evaluation of 3D cine images demonstrates that CORe is more effective in suppressing artifacts while maintaining or improving image sharpness. Finally, 4D flow images show that CORe yields more reliable and consistent flow measurements, especially in the presence of involuntary subject motion or exercise stress., Conclusion: An outlier rejection method is presented and tested using simulated and measured data. This method can help suppress motion artifacts in a wide range of free-running CMR applications., (© 2024 The Authors. Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.)

Adapting a medical image segmentation model to a new domain is important for improving its cross-domain transferability, and due to the expensive annotation process, Unsupervised Domain Adaptation (UDA) is appealing where only unlabeled images are needed for the adaptation. Existing UDA methods are mainly based on image or feature alignment with adversarial training for regularization, and they are limited by insufficient supervision in the target domain. In this paper, we propose an enhanced Filtered Pseudo Label (FPL+)-based UDA method for 3D medical image segmentation. It first uses cross-domain data augmentation to translate labeled images in the source domain to a dual-domain training set consisting of a pseudo source-domain set and a pseudo target-domain set. To leverage the dual-domain augmented images to train a pseudo label generator, domain-specific batch normalization layers are used to deal with the domain shift while learning the domain-invariant structure features, generating high-quality pseudo labels for target-domain images. We then combine labeled source-domain images and target-domain images with pseudo labels to train a final segmentor, where image-level weighting based on uncertainty estimation and pixel-level weighting based on dual-domain consensus are proposed to mitigate the adverse effect of noisy pseudo labels. Experiments on three public multi-modal datasets for Vestibular Schwannoma, brain tumor and whole heart segmentation show that our method surpassed ten state-of-the-art UDA methods, and it even achieved better results than fully supervised learning in the target domain in some cases.

Owing to the success of transformer models, recent works study their applicability in 3D medical segmentation tasks. Within the transformer models, the self-attention mechanism is one of the main building blocks that strives to capture long-range dependencies, compared to the local convolutional-based design. However, the self-attention operation has quadratic complexity which proves to be a computational bottleneck, especially in volumetric medical imaging, where the inputs are 3D with numerous slices. In this paper, we propose a 3D medical image segmentation approach, named UNETR++, that offers both high-quality segmentation masks as well as efficiency in terms of parameters, compute cost, and inference speed. The core of our design is the introduction of a novel efficient paired attention (EPA) block that efficiently learns spatial and channel-wise discriminative features using a pair of inter-dependent branches based on spatial and channel attention. Our spatial attention formulation is efficient and has linear complexity with respect to the input. To enable communication between spatial and channel-focused branches, we share the weights of query and key mapping functions that provide a complimentary benefit (paired attention), while also reducing the complexity. Our extensive evaluations on five benchmarks, Synapse, BTCV, ACDC, BraTS, and Decathlon-Lung, reveal the effectiveness of our contributions in terms of both efficiency and accuracy. On Synapse, our UNETR++ sets a new state-of-the-art with a Dice Score of 87.2%, while significantly reducing parameters and FLOPs by over 71%, compared to the best method in the literature. Our code and models are available at: https://tinyurl.com/2p87x5xn.

Purpose: MRI using 3D stack-of-spirals (SoS) readout on a high-performance gradient system is subject to strong second-order, spatially varying concomitant fields, which can lead to signal dropout and blurring artifacts that become more significant at locations farther from the gradient isocenter. A method for compensating for second-order concomitant fields in 3D axial SoS image reconstruction is described., Methods: We retrospectively correct for second-order concomitant field-induced phase error in the 3D SoS data by slice-dependent k-space phase compensation based on the nominal spiral readout trajectories. The effectiveness of the method was demonstrated in phantom and healthy volunteer scans in which 3D pseudo-continuous arterial spin labeling imaging was performed with SoS fast spin-echo readout at 3 T., Results: Substantial reduction in blurring was observed with the proposed method. In phantom scans, blurring was reduced by about 53% at 98 mm from the gradient isocenter. In the in vivo 3D pseudo-continuous arterial spin labeling scans, differences of up to 10% were observed at 78 mm from the isocenter, especially around the white-matter and gray-matter interfaces, between the corrected and uncorrected proton density images, perfusion-weighted images, and cerebral blood flow maps., Conclusions: The described retrospective correction method provides a means to correct erroneous phase accruals due to second-order concomitant fields in 3D axial stack-of-spirals imaging., (© 2024 International Society for Magnetic Resonance in Medicine.)

Background: For bone morphology and biomechanics analysis, landmarks are essential to define position, orientation, and shape. These landmarks define bone and joint coordinate systems and are widely used in these research fields. Currently, no method is known for automatically identifying landmarks on virtual 3D bone models of the radius and ulna. This paper proposes a knowledge-based method for locating landmarks and calculating a coordinate system for the radius, ulna, and combined forearm bones, which is essential for measuring forearm function. This method does not rely on pre-labeled data., Validation: The algorithm is validated by comparing the landmarks placed by the algorithm with the mean position of landmarks placed by a group of experts on cadaveric specimens regarding distance and orientation., Results: The median Euclidean distance differences between all the automated and reference landmarks range from 0.4 to 1.8 millimeters. The median angular differences of the coordinate system of the radius and ulna range from -1.4 to 0.6 degrees. The forearm coordinate system's median errors range from -0.2 to 2.0 degrees. The median error in calculating the rotational position of the radius relative to the ulna is 1.8 degrees., Conclusion: The automatic method's applicability depends on the use context and desired accuracy. However, the current method is a validated first step in the automatic analysis of the three-dimensional forearm anatomy., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 The Author(s). Published by Elsevier Ltd.. All rights reserved.)

Objectives: Severity of degenerative scoliosis (DS) is assessed by measuring the Cobb angle on anteroposterior radiographs. However, MRI images are often available to study the degenerative spine. This retrospective study aims to develop and evaluate the reliability of a novel automatic method that measures coronal Cobb angles on lumbar MRI in DS patients., Materials and Methods: Vertebrae and intervertebral discs were automatically segmented using a 3D AI algorithm, trained on 447 lumbar MRI series. The segmentations were used to calculate all possible angles between the vertebral endplates, with the largest being the Cobb angle. The results were validated with 50 high-resolution sagittal lumbar MRI scans of DS patients, in which three experienced readers measured the Cobb angle. Reliability was determined using the intraclass correlation coefficient (ICC)., Results: The ICCs between the readers ranged from 0.90 (95% CI 0.83-0.94) to 0.93 (95% CI 0.88-0.96). The ICC between the maximum angle found by the algorithm and the average manually measured Cobb angles was 0.83 (95% CI 0.71-0.90). In 9 out of the 50 cases (18%), all readers agreed on both vertebral levels for Cobb angle measurement. When using the algorithm to extract the angles at the vertebral levels chosen by the readers, the ICCs ranged from 0.92 (95% CI 0.87-0.96) to 0.97 (95% CI 0.94-0.98)., Conclusion: The Cobb angle can be accurately measured on MRI using the newly developed algorithm in patients with DS. The readers failed to consistently choose the same vertebral level for Cobb angle measurement, whereas the automatic approach ensures the maximum angle is consistently measured., Clinical Relevance Statement: Our AI-based algorithm offers reliable Cobb angle measurement on routine MRI for degenerative scoliosis patients, potentially reducing the reliance on conventional radiographs, ensuring consistent assessments, and therefore improving patient care., Key Points: • While often available, MRI images are rarely utilized to determine the severity of degenerative scoliosis. • The presented MRI Cobb angle algorithm is more reliable than humans in patients with degenerative scoliosis. • Radiographic imaging for Cobb angle measurements is mitigated when lumbar MRI images are available., (© 2024. The Author(s).)

Purpose: To study the additional value of FRONSAC encoding in 2D and 3D wave sequences, implementing a simple strategy to trajectory mapping for FRONSAC encoding gradients., Theory and Methods: The nonlinear gradient trajectory for each voxel was estimated by exploiting the sparsity of the point spread function in the frequency domain. Simulations and in-vivo experiments were used to analyze the performance of combinations of wave and FRONSAC encoding., Results: Field mapping using the simplified approach produced similar image quality with much shorter calibration time than the comprehensive mapping schemes utilized in previous work. In-vivo human brain images showed that the addition of FRONSAC encoding could improve wave image quality, particularly at very high undersampling factors and in the context of limited wave amplitudes. These results were further supported by g-factor maps., Conclusion: Results show that FRONSAC can be used to improve image quality of wave at very high undersampling rates or in slew-limited acquisitions. Our study illustrates the potential of the proposed fast field mapping approach., (© 2024 International Society for Magnetic Resonance in Medicine.)

In the field of medical imaging, the fusion of data from diverse modalities plays a pivotal role in advancing our understanding of pathological conditions. Sparse representation (SR), a robust signal modeling technique, has demonstrated noteworthy success in multi-dimensional (MD) medical image fusion. However, a fundamental limitation appearing in existing SR models is their lack of directionality, restricting their efficacy in extracting anatomical details from different imaging modalities. To tackle this issue, we propose a novel directional SR model, termed complex sparse representation (ComSR), specifically designed for medical image fusion. ComSR independently represents MD signals over directional dictionaries along specific directions, allowing precise analysis of intricate details of MD signals. Besides, current studies in medical image fusion mostly concentrate on addressing either 2D or 3D fusion problems. This work bridges this gap by proposing a MD medical image fusion method based on ComSR, presenting a unified framework for both 2D and 3D fusion tasks. Experimental results across six multi-modal medical image fusion tasks, involving 93 pairs of 2D source images and 20 pairs of 3D source images, substantiate the superiority of our proposed method over 11 state-of-the-art 2D fusion methods and 4 representative 3D fusion methods, in terms of both visual quality and objective evaluation.

Robust segmentation of large and complex conjoined tree structures in 3-D is a major challenge in computer vision. This is particularly true in computational biology, where we often encounter large data structures in size, but few in number, which poses a hard problem for learning algorithms. We show that merging multiscale opening with geodesic path propagation, can shed new light on this classic machine vision challenge, while circumventing the learning issue by developing an unsupervised visual geometry approach (digital topology/morphometry). The novelty of the proposed MSO-GP method comes from the geodesic path propagation being guided by a skeletonization of the conjoined structure that helps to achieve robust segmentation results in a particularly challenging task in this area, that of artery-vein separation from non-contrast pulmonary computed tomography angiograms. This is an important first step in measuring vascular geometry to then diagnose pulmonary diseases and to develop image-based phenotypes. We first present proof-of-concept results on synthetic data, and then verify the performance on pig lung and human lung data with less segmentation time and user intervention needs than those of the competing methods., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier Inc. All rights reserved.)

Purpose: An automatic method is presented for estimating 4D flow MRI velocity measurement uncertainty in each voxel. The velocity distance (VD) metric, a statistical distance between the measured velocity and local error distribution, is introduced as a novel measure of 4D flow MRI velocity measurement quality., Methods: The method uses mass conservation to assess the local velocity error variance and the standardized difference of means (SDM) velocity to estimate the velocity error correlations. VD is evaluated as the Mahalanobis distance between the local velocity measurement and the local error distribution. The uncertainty model is validated synthetically and tested in vitro under different flow resolutions and noise levels. The VD's application is demonstrated on two in vivo thoracic vasculature 4D flow datasets., Results: Synthetic results show the proposed uncertainty quantification method is sensitive to aliased regions across various velocity-to-noise ratios and assesses velocity error correlations in four- and six-point acquisitions with correlation errors at or under 3.2%. In vitro results demonstrate the method's sensitivity to spatial resolution, venc settings, partial volume effects, and phase wrapping error sources. Applying VD to assess in vivo 4D flow MRI in the aorta demonstrates the expected increase in measured velocity quality with contrast administration and systolic flow., Conclusion: The proposed 4D flow MRI uncertainty quantification method assesses velocity measurement error owing to sources including noise, intravoxel phase dispersion, and velocity aliasing. This method enables rigorous comparison of 4D flow MRI datasets obtained in longitudinal studies, across patient populations, and with different MRI systems., (© 2024 The Author(s). Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.)

Purpose: MRI-guidance of cardiac catheterization is currently performed using one or multiple 2D imaging planes, which may be suboptimal for catheter navigation, especially in patients with complex anatomies. The purpose of the work was to develop a robust real-time 3D catheter tracking method and 3D visualization strategy for improved MRI-guidance of cardiac catheterization procedures., Methods: A fast 3D tracking technique was developed using continuous acquisition of two orthogonal 2D-projection images. Each projection corresponds to a gradient echo stack of slices with only the central k-space lines being collected for each slice. To enhance catheter contrast, a saturation pulse is added ahead of the projection pair. An offline image processing algorithm was developed to identify the 2D coordinates of the balloon in each projection image and to estimate its corresponding 3D coordinates. Post-processing includes background signal suppression using an atlas of background 2D-projection images. 3D visualization of the catheter and anatomy is proposed using three live sagittal, coronal, and axial (MPR) views and 3D rendering. The technique was tested in a subset of a catheterization step in three patients undergoing MRI-guided cardiac catheterization using a passive balloon catheter., Results: The extraction of the catheter balloon 3D coordinates was successful in all patients and for the majority of time-points (accuracy >96%). This tracking method enabled a novel 3D visualization strategy for passive balloon catheter, providing enhanced anatomical context during catheter navigation., Conclusion: The proposed tracking strategy shows promise for robust tracking of passive balloon catheter and may enable enhanced visualization during MRI-guided cardiac catheterization., (© 2024 The Author(s). Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.)

Purpose: To provide a navigator-based run-time motion and first-order field correction for three-dimensional human brain imaging with high precision, minimal calibration and acquisition, and fast processing., Methods: A complex-valued linear perturbation model with feedback control is extended to estimate and correct for gradient shim fields using orbital navigators (2.3 ms). Two approaches for sensitizing the model to gradient fields are presented, one based on finite differences with three additional navigators, and another projection-based approximation requiring no additional navigators. A mechanism for noise decorrelation of the matrix and the data is proposed and evaluated to reduce unwanted parameter biases., Results: The rigid motion and first-order field control achieves robust motion and gradient shim corrections improving image quality in a series of phantom and in vivo experiments with varying field conditions. In phantom scans, magnet drifts, forced gradient field perturbations and field distortions from shifts of a second bottle phantom are successfully corrected. Field estimates of the magnet drifts are in good agreement with concurrent field probe measurements. For in vivo scans, the proposed method mitigates field variations from torso motions while being robust to head motion. In vivo gradient field precisions were 30   nT / m $$ 30\;\mathrm{nT}/\mathrm{m} $$ along with single-digit micrometer and millidegree rigid precisions., Conclusion: The navigator-based method achieves accurate, high-precision run-time motion and field corrections with low sequence impact and calibration requirements., (© 2024 The Author(s). Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.)

Purpose: Data for QSM are typically acquired using multi-echo 3D gradient echo (GRE), but EPI can be used to accelerate QSM and provide shorter acquisition times. So far, EPI-QSM has been limited to single-echo acquisitions, which, for 3D GRE, are known to be less accurate than multi-echo sequences. Therefore, we compared single-echo and multi-echo EPI-QSM reconstructions across a range of parallel imaging and multiband acceleration factors., Methods: Using 2D single-shot EPI in the brain, we compared QSM from single-echo and multi-echo acquisitions across combined parallel-imaging and multiband acceleration factors ranging from 2 to 16, with volume pulse TRs from 21.7 to 3.2 s, respectively. For single-echo versus multi-echo reconstructions, we investigated the effect of acceleration factors on regional susceptibility values, temporal noise, and image quality. We introduce a novel masking method based on thresholding the magnitude of the local field gradients to improve brain masking in challenging regions., Results: At 1.6-mm isotropic resolution, high-quality QSM was achieved using multi-echo 2D EPI with a combined acceleration factor of 16 and a TR of 3.2 s, which enables functional applications. With these high acceleration factors, single-echo reconstructions are inaccurate and artefacted, rendering them unusable. Multi-echo acquisitions greatly improve QSM quality, particularly at higher acceleration factors, provide more consistent regional susceptibility values across acceleration factors, and decrease temporal noise compared with single-echo QSM reconstructions., Conclusion: Multi-echo acquisition is more robust for EPI-QSM across parallel imaging and multiband acceleration factors than single-echo acquisition. Multi-echo EPI can be used for highly accelerated acquisition while preserving QSM accuracy and quality relative to gold-standard 3D-GRE QSM., (© 2024 The Author(s). Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.)

Background and Objectives: Image-based 2D/3D registration is a crucial technology for fluoroscopy-guided surgical interventions. However, traditional registration methods relying on a single X-ray image into surgical navigation systems. This study proposes a novel 2D/3D registration approach utilizing biplanar X-ray images combined with computed tomography (CT) to significantly reduce registration and navigation errors. The method is successfully implemented in a surgical navigation system, enhancing its precision and reliability., Methods: First, we simultaneously register the frontal and lateral X-ray images with the CT image, enabling mutual complementation and more precise localization. Additionally, we introduce a novel similarity measure for image comparison, providing a more robust cost function for the optimization algorithm. Furthermore, a multi-resolution strategy is employed to enhance registration efficiency. Lastly, we propose a more accurate coordinate transformation method, based on projection and 3D reconstruction, to improve the precision of surgical navigation systems., Results: We conducted registration and navigation experiments using pelvic, spinal, and femur phantoms. The navigation results demonstrated that the feature registration errors (FREs) in the three experiments were 0.505±0.063 mm, 0.515±0.055 mm, and 0.577±0.056 mm, respectively. Compared to the point-to-point (PTP) registration method based on anatomical landmarks, our method reduced registration errors by 31.3%, 23.9%, and 26.3%, respectively., Conclusion: The results demonstrate that our method significantly reduces registration and navigation errors, highlighting its potential for application across various anatomical sites. Our code is available at: https://github.com/SJTUdemon/2D-3D-Registration., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier B.V. All rights reserved.)

Stereoscopic images typically consist of left and right views along with depth information. Assessing the quality of stereoscopic/3D images (SIQA) is often more complex than that of 2D images due to scene disparities between the left and right views and the intricate process of fusion in binocular vision. To address the problem of quality prediction bias of multi-distortion images, we investigated the visual physiology and the processing of visual information by the primary visual cortex of the human brain and proposed a no-reference stereoscopic image quality evaluation method. The method mainly includes an innovative end-to-end NR-SIQA neural network with a picture patch generation algorithm. The algorithm generates a saliency map by fusing the left and right views and then guides the image cropping in the database based on the saliency map. The proposed models are validated and compared based on publicly available databases. The results show that the model and algorithm together outperform the state-of-the-art NR-SIQA metric in the LIVE 3D database and the WIVC 3D database, and have excellent results in the specific noise metric. The model generalization experiments demonstrate a certain degree of generality of our proposed model., Competing Interests: Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Wenzhong Guo reports financial support was provided by National Natural Science Foundation of China. Wenzhong Guo reports financial support was provided by National Key Research and Development Plan of China. Wenzhong Guo reports was provided by Natural Science Foundation of Fujian Province. Wenzhong Guo reports was provided by Major Science and Technology Project of Fujian Province. Wenzhong Guo reports financial support was provided by Industry-Academy Cooperation Project of Fujian Province. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier Ltd. All rights reserved.)

Tumors are an important health concern in modern times. Breast cancer is one of the most prevalent causes of death for women. Breast cancer is rapidly becoming the leading cause of mortality among women globally. Early detection of breast cancer allows patients to obtain appropriate therapy, increasing their probability of survival. The adoption of 3-Dimensional (3D) mammography for the medical identification of abnormalities in the breast reduced the number of deaths dramatically. Classification and accurate detection of lumps in the breast in 3D mammography is especially difficult due to factors such as inadequate contrast and normal fluctuations in tissue density. Several Computer-Aided Diagnosis (CAD) solutions are under development to help radiologists accurately classify abnormalities in the breast. In this paper, a breast cancer diagnosis model is implemented to detect breast cancer in cancer patients to prevent death rates. The 3D mammogram images are gathered from the internet. Then, the gathered images are given to the preprocessing phase. The preprocessing is done using a median filter and image scaling method. The purpose of the preprocessing phase is to enhance the quality of the images and remove any noise or artifacts that may interfere with the detection of abnormalities. The median filter helps to smooth out any irregularities in the images, while the image scaling method adjusts the size and resolution of the images for better analysis. Once the preprocessing is complete, the preprocessed image is given to the segmentation phase. The segmentation phase is crucial in medical image analysis as it helps to identify and separate different structures within the image, such as organs or tumors. This process involves dividing the preprocessed image into meaningful regions or segments based on intensity, color, texture, or other features. The segmentation process is done using Adaptive Thresholding with Region Growing Fusion Model (AT-RGFM)". This model combines the advantages of both thresholding and region-growing techniques to accurately identify and delineate specific structures within the image. By utilizing AT-RGFM, the segmentation phase can effectively differentiate between different parts of the image, allowing for more precise analysis and diagnosis. It plays a vital role in the medical image analysis process, providing crucial insights for healthcare professionals. Here, the Modified Garter Snake Optimization Algorithm (MGSOA) is used to optimize the parameters. It helps to optimize parameters for accurately identifying and delineating specific structures within medical images and also helps healthcare professionals in providing more precise analysis and diagnosis, ultimately playing a vital role in the medical image analysis process. MGSOA enhances the segmentation phase by effectively differentiating between different parts of the image, leading to more accurate results. Then, the segmented image is fed into the detection phase. The tumor detection is performed by the Vision Transformer-based Multiscale Adaptive EfficientNetB7 (ViT-MAENB7) model. This model utilizes a combination of advanced algorithms and deep learning techniques to accurately identify and locate tumors within the segmented medical image. By incorporating a multiscale adaptive approach, the ViT-MAENB7 model can analyze the image at various levels of detail, improving the overall accuracy of tumor detection. This crucial step in the medical image analysis process allows healthcare professionals to make more informed decisions regarding patient treatment and care. Here, the created MGSOA algorithm is used to optimize the parameters for enhancing the performance of the model. The suggested breast cancer diagnosis performance is compared to conventional cancer diagnosis models and it showed high accuracy. The accuracy of the developed MGSOA-ViT-MAENB7 is 96.6 %, and others model like RNN, LSTM, EffNet, and ViT-MAENet given the accuracy to be 90.31 %, 92.79 %, 94.46 % and 94.75 %. The developed model's ability to analyze images at multiple scales, combined with the optimization provided by the MGSOA algorithm, results in a highly accurate and efficient system for detecting tumors in medical images. This cutting-edge technology not only improves the accuracy of diagnosis but also helps healthcare professionals tailor treatment plans to individual patients, ultimately leading to better outcomes. By outperforming traditional cancer diagnosis models, the proposed model is revolutionizing the field of medical imaging and setting a new standard for precision and effectiveness in healthcare., Competing Interests: Declaration of competing interest The authors declare no conflict of interest., (Copyright © 2024 Elsevier B.V. All rights reserved.)

Quantitative MRI utilizes multiple acquisitions with varying sequence parameters to sufficiently characterize a biophysical model of interest, resulting in undesirable scan times. Here we propose, validate and demonstrate a new general strategy for accelerating MRI using subvoxel shifting as a source of encoding called POSition Encoding (POSE). The POSE framework applies unique subvoxel shifts along the acquisition parameter dimension, thereby creating an extra source of encoding. Combining with a biophysical signal model of interest, accelerated and enhanced resolution maps of biophysical parameters are obtained. This has been validated and demonstrated through numerical Bloch equation simulations, phantom experiments and in vivo experiments using the variable flip angle signal model in 3D acquisitions as an application example. Monte Carlo simulations were performed using in vivo data to investigate our method's noise performance. POSE quantification results from numerical Bloch equation simulations of both a numerical phantom and realistic digital brain phantom concur well with the reference method, validating our method both theoretically and for realistic situations. NIST phantom experiment results show excellent overall agreement with the reference method, confirming our method's applicability for a wide range of T 1 values. In vivo results not only exhibit good agreement with the reference method, but also show g-factors that significantly outperforms conventional parallel imaging methods with identical acceleration. Furthermore, our results show that POSE can be combined with parallel imaging to further accelerate while maintaining superior noise performance over parallel imaging that uses lower acceleration factors., Competing Interests: Declaration of competing interest The authors have no conflict of interest to declare., (Copyright © 2024 Elsevier Inc. All rights reserved.)

X-ray Computed Tomography (CT) images are widely used in various fields of natural, physical, and biological sciences. 3D reconstruction of the images involves segmentation of the structures of interest. Manual segmentation has been widely used in the field of biological sciences for complex structures composed of several sub-parts and can be a time-consuming process. Many tools have been developed to automate the segmentation process, all with various limitations and advantages, however, multipart segmentation remains a largely manual process. The aim of this study was to develop an open-access and user-friendly tool for the automatic segmentation of calcified tissues, specifically focusing on craniofacial bones. Here we describe BounTI, a novel segmentation algorithm which preserves boundaries between separate segments through iterative thresholding. This study outlines the working principles behind this algorithm, investigates the effect of several input parameters on its outcome, and then tests its versatility on CT images of the craniofacial system from different species (e.g. a snake, a lizard, an amphibian, a mouse and a human skull) with various scan qualities. The case studies demonstrate that this algorithm can be effectively used to segment the craniofacial system of a range of species automatically. High-resolution microCT images resulted in more accurate boundary-preserved segmentation, nonetheless significantly lower-quality clinical images could still be segmented using the proposed algorithm. Methods for manual intervention are included in this tool when the scan quality is insufficient to achieve the desired segmentation results. While the focus here was on the craniofacial system, BounTI can be used to automatically segment any hard tissue. The tool presented here is available as an Avizo/Amira add-on, a stand-alone Windows executable, and a Python library. We believe this accessible and user-friendly segmentation tool can benefit the wider anatomical community., (© 2024 The Authors. Journal of Anatomy published by John Wiley & Sons Ltd on behalf of Anatomical Society.)

The Segment Anything Model (SAM), a foundation model for general image segmentation, has demonstrated impressive zero-shot performance across numerous natural image segmentation tasks. However, SAM's performance significantly declines when applied to medical images, primarily due to the substantial disparity between natural and medical image domains. To effectively adapt SAM to medical images, it is important to incorporate critical third-dimensional information, i.e., volumetric or temporal knowledge, during fine-tuning. Simultaneously, we aim to harness SAM's pre-trained weights within its original 2D backbone to the fullest extent. In this paper, we introduce a modality-agnostic SAM adaptation framework, named as MA-SAM, that is applicable to various volumetric and video medical data. Our method roots in the parameter-efficient fine-tuning strategy to update only a small portion of weight increments while preserving the majority of SAM's pre-trained weights. By injecting a series of 3D adapters into the transformer blocks of the image encoder, our method enables the pre-trained 2D backbone to extract third-dimensional information from input data. We comprehensively evaluate our method on five medical image segmentation tasks, by using 11 public datasets across CT, MRI, and surgical video data. Remarkably, without using any prompt, our method consistently outperforms various state-of-the-art 3D approaches, surpassing nnU-Net by 0.9%, 2.6%, and 9.9% in Dice for CT multi-organ segmentation, MRI prostate segmentation, and surgical scene segmentation respectively. Our model also demonstrates strong generalization, and excels in challenging tumor segmentation when prompts are used. Our code is available at: https://github.com/cchen-cc/MA-SAM., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier B.V. All rights reserved.)

In this paper, we propose a method to reduce the model architecture searching time. We consider MobileNetV2 for 3D face recognition tasks as a case study and introducing the layer replication to enhance accuracy. For a given network, various layers can be replicated, and effective replication can yield better accuracy. Our proposed algorithm identifies the optimal layer replication configuration for the model. We considered two acceleration methods: distributed data-parallel training and concurrent model training. Our experiments demonstrate the effectiveness of the automatic model finding process for layer replication, using both distributed data-parallel and concurrent training under different conditions. The accuracy of our model improved by up to 6% compared to the previous work on 3D MobileNetV2, and by 8% compared to the vanilla MobileNetV2. Training models with distributed data-parallel across four GPUs reduced model training time by up to 75% compared to traditional training on a single GPU. Additionally, the automatic model finding process with concurrent training was 1,932 minutes faster than the distributed training approach in finding an optimal solution., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2024 Soongswang, Chantrapornchai. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)