Search

Abstract Numerous deep-learning models have been developed using task-specific data, but they ignore the inherent connections among different tasks. By jointly learning a wide range of segmentation tasks, we prove that a general medical image segmentation model can improve segmentation performance for computerized tomography (CT) volumes. The proposed general CT image segmentation (gCIS) model utilizes a common transformer-based encoder for all tasks and incorporates automatic pathway modules for task prompt-based decoding. It is trained on one of the largest datasets, comprising 36,419 CT scans and 83 tasks. gCIS can automatically perform various segmentation tasks using automatic pathway modules of decoding networks through text prompt inputs, achieving an average Dice coefficient of 82.84%. Furthermore, the proposed automatic pathway routing mechanism allows for parameter pruning of the network during deployment, and gCIS can also be quickly adapted to unseen tasks with minimal training samples while maintaining great performance.

Abstract Computer-aided diagnosis (CAD) has advanced medical image analysis, while large language models (LLMs) have shown potential in clinical applications. However, LLMs struggle to interpret medical images, which are critical for decision-making. Here we show a strategy integrating LLMs with CAD networks. The framework uses LLMs’ medical knowledge and reasoning to enhance CAD network outputs, such as diagnosis, lesion segmentation, and report generation, by summarizing information in natural language. The generated reports are of higher quality and can improve the performance of vision-based CAD models. In chest X-rays, an LLM using ChatGPT improved diagnosis performance by 16.42 percentage points compared to state-of-the-art models, while GPT-3 provided a 15.00 percentage point F1-score improvement. Our strategy allows accurate report generation and creates a patient-friendly interactive system, unlike conventional CAD systems only understood by professionals. This approach has the potential to revolutionize clinical decision-making and patient communication.

Abstract Brain disease diagnosis using multiple imaging modalities has shown superior performance compared to using single modality, yet multi-modal data is not easily available in clinical routine due to cost or radiation risk. Here we propose a synthesis-empowered uncertainty-aware classification framework for brain disease diagnosis. To synthesize disease-relevant features effectively, a two-stage framework is proposed including multi-modal feature representation learning and representation transfer based on hierarchical similarity matching. Besides, the synthesized and acquired modality features are integrated based on evidential learning, which provides diagnosis decision and also diagnosis uncertainty. Our framework is extensively evaluated on five datasets containing 3758 subjects for three brain diseases including Alzheimer’s disease (AD), subcortical vascular mild cognitive impairment (MCI), and O[6]-methylguanine-DNA methyltransferase promoter methylation status for glioblastoma, achieving 0.950 and 0.806 in area under the ROC curve on ADNI dataset for discriminating AD patients from normal controls and progressive MCI from static MCI, respectively. Our framework not only achieves quasi-multimodal performance although using single-modal input, but also provides reliable diagnosis uncertainty.

Abstract Background Though deep learning has consistently demonstrated advantages in the automatic interpretation of breast ultrasound images, its black-box nature hinders potential interactions with radiologists, posing obstacles for clinical deployment. Methods We proposed a domain knowledge-based interpretable deep learning system for improving breast cancer risk prediction via paired multimodal ultrasound images. The deep learning system was developed on 4320 multimodal breast ultrasound images of 1440 biopsy-confirmed lesions from 1348 prospectively enrolled patients across two hospitals between August 2019 and December 2022. The lesions were allocated to 70% training cohort, 10% validation cohort, and 20% test cohort based on case recruitment date. Results Here, we show that the interpretable deep learning system can predict breast cancer risk as accurately as experienced radiologists, with an area under the receiver operating characteristic curve of 0.902 (95% confidence interval = 0.882 – 0.921), sensitivity of 75.2%, and specificity of 91.8% on the test cohort. With the aid of the deep learning system, particularly its inherent explainable features, junior radiologists tend to achieve better clinical outcomes, while senior radiologists experience increased confidence levels. Multimodal ultrasound images augmented with domain knowledge-based reasoning cues enable an effective human-machine collaboration at a high level of prediction performance. Conclusions Such a clinically applicable deep learning system may be incorporated into future breast cancer screening and support assisted or second-read workflows.

BackgroundLarge language models show promise for improving radiology workflows, but their performance on structured radiological tasks such as Reporting and Data Systems (RADS) categorization remains unexplored. ObjectiveThis study aims to evaluate 3 large language model chatbots—Claude-2, GPT-3.5, and GPT-4—on assigning RADS categories to radiology reports and assess the impact of different prompting strategies. MethodsThis cross-sectional study compared 3 chatbots using 30 radiology reports (10 per RADS criteria), using a 3-level prompting strategy: zero-shot, few-shot, and guideline PDF-informed prompts. The cases were grounded in Liver Imaging Reporting & Data System (LI-RADS) version 2018, Lung CT (computed tomography) Screening Reporting & Data System (Lung-RADS) version 2022, and Ovarian-Adnexal Reporting & Data System (O-RADS) magnetic resonance imaging, meticulously prepared by board-certified radiologists. Each report underwent 6 assessments. Two blinded reviewers assessed the chatbots’ response at patient-level RADS categorization and overall ratings. The agreement across repetitions was assessed using Fleiss κ. ResultsClaude-2 achieved the highest accuracy in overall ratings with few-shot prompts and guideline PDFs (prompt-2), attaining 57% (17/30) average accuracy over 6 runs and 50% (15/30) accuracy with k-pass voting. Without prompt engineering, all chatbots performed poorly. The introduction of a structured exemplar prompt (prompt-1) increased the accuracy of overall ratings for all chatbots. Providing prompt-2 further improved Claude-2’s performance, an enhancement not replicated by GPT-4. The interrun agreement was substantial for Claude-2 (k=0.66 for overall rating and k=0.69 for RADS categorization), fair for GPT-4 (k=0.39 for both), and fair for GPT-3.5 (k=0.21 for overall rating and k=0.39 for RADS categorization). All chatbots showed significantly higher accuracy with LI-RADS version 2018 than with Lung-RADS version 2022 and O-RADS (P

At the dawn of of Artificial General Intelligence (AGI), the emergence of large language models such as ChatGPT show promise in revolutionizing healthcare by improving patient care, expanding medical access, and optimizing clinical processes. However, their integration into healthcare systems requires careful consideration of potential risks, such as inaccurate medical advice, patient privacy violations, the creation of falsified documents or images, overreliance on AGI in medical education, and the perpetuation of biases. It is crucial to implement proper oversight and regulation to address these risks, ensuring the safe and effective incorporation of AGI technologies into healthcare systems. By acknowledging and mitigating these challenges, AGI can be harnessed to enhance patient care, medical knowledge, and healthcare processes, ultimately benefiting society as a whole.

Abstract Medical image reconstruction and synthesis are critical for imaging quality, disease diagnosis and treatment. Most of the existing generative models ignore the fact that medical imaging usually occurs in the acquisition domain, which is different from, but associated with, the image domain. Such methods exploit either single-domain or dual-domain information and suffer from inefficient information coupling across domains. Moreover, these models are usually designed specifically and not general enough for different tasks. Here we present a generalized dual-domain generative framework to facilitate the connections within and across domains by elaborately-designed hierarchical consistency constraints. A multi-stage learning strategy is proposed to construct hierarchical constraints effectively and stably. We conducted experiments for representative generative tasks including low-dose PET/CT reconstruction, CT metal artifact reduction, fast MRI reconstruction, and PET/CT synthesis. All these tasks share the same framework and achieve better performance, which validates the effectiveness of our framework. This technology is expected to be applied in clinical imaging to increase diagnosis efficiency and accuracy.

Summary: Accurate pathological classification and grading of gliomas is crucial in clinical diagnosis and treatment. The application of deep learning techniques holds promise for automated histological pathology diagnosis. In this study, we collected 733 whole slide images from four medical centers, of which 456 were used for model training, 150 for internal validation, and 127 for multi-center testing. The study includes 5 types of common gliomas.A subtask-guided multi-instance learning image-to-label training pipeline was employed. The pipeline leveraged “patch prompting” for the model to converge with reasonable computational cost. Experiments showed that an overall accuracy of 0.79 in the internal validation dataset. The performance on the multi-center testing dataset showed an overall accuracy to 0.73. The findings suggest a minor yet acceptable performance decrease in multi-center data, demonstrating the model’s strong generalizability and establishing a robust foundation for future clinical applications.

Abstract Visual interpretation is considered the gold standard for amyloid scans in clinical practice. However, dichotomous classification of amyloid deposition by visual reading always results in bias due to rater experience. Therefore, there is a need for a more lenient and flexible amyloid‐equivocal classification in clinical practice. A total of 461 participants were included in this study. Amyloid and glucose positron‐emission tomography was performed, and neuropsychological tests were evaluated. A disease‐specific deep‐learning method was used to identify amyloid equivocality. Amyloid deposition, glucose metabolism, and cognitive performance were analyzed and compared among amyloid‐positive, amyloid‐negative, and amyloid‐equivocal groups. Clinically diagnosed Alzheimer's disease individuals and subjects with normal cognition were used to create amyloid biomarker cut points to support the definition of equivocal amyloid deposition. A total of 139 amyloid‐equivocal individuals were identified by deep learning methods. They displayed intermediate amyloid deposition between that of amyloid‐positive (standardized uptake value ratio [SUVr]: 1.25 ± 0.10 vs. 1.47 ± 0.20, p < 0.001) and amyloid‐negative (SUVr: 1.25 ± 0.10 vs. 1.18 ± 0.07, p < 0.001) individuals. No difference in glucose metabolism or cognitive performance was observed between amyloid negativity and equivocality. Furthermore, the SUVr for the whole cortex, the precuneus, and the frontal lobe served as auxiliary criteria supporting the diagnosis of equivocal amyloid deposition. We also established a guide to assist in the interpretation of amyloid equivocality by visual reading with auxiliary criteria including two cut points and deep learning methods.

Human brain undergoes rapid growth during the first few years of life. While previous research has employed graph theory to study early brain development, it has mostly focused on the topological attributes of the whole brain. However, examining regional graph-theory features may provide unique insights into the development of cognitive abilities. Utilizing a large and longitudinal rsfMRI dataset from the UNC/UMN Baby Connectome Project, we investigated the developmental trajectories of regional efficiency and evaluated the relationships between these changes and cognitive abilities using Mullen Scales of Early Learning during the first twenty-eight months of life. Our results revealed a complex and spatiotemporally heterogeneous development pattern of regional global and local efficiency during this age period. Furthermore, we found that the trajectories of the regional global efficiency at the left temporal occipital fusiform and bilateral occipital fusiform gyri were positively associated with cognitive abilities, including visual reception, expressive language, receptive language, and early learning composite scores (P

Volume delineation of organs-at risk (OARs) and target tumors is an indispensable process for creating radiotherapy treatment planning. Herein, the authors propose a lightweight deep learning framework to empower the rapid and precise volume delineation of whole-body OARs and target tumors.

IntroductionMedical image analysis is of tremendous importance in serving clinical diagnosis, treatment planning, as well as prognosis assessment. However, the image analysis process usually involves multiple modality-specific software and relies on rigorous manual operations, which is time-consuming and potentially low reproducible.MethodsWe present an integrated platform - uAI Research Portal (uRP), to achieve one-stop analyses of multimodal images such as CT, MRI, and PET for clinical research applications. The proposed uRP adopts a modularized architecture to be multifunctional, extensible, and customizable.Results and DiscussionThe uRP shows 3 advantages, as it 1) spans a wealth of algorithms for image processing including semi-automatic delineation, automatic segmentation, registration, classification, quantitative analysis, and image visualization, to realize a one-stop analytic pipeline, 2) integrates a variety of functional modules, which can be directly applied, combined, or customized for specific application domains, such as brain, pneumonia, and knee joint analyses, 3) enables full-stack analysis of one disease, including diagnosis, treatment planning, and prognosis assessment, as well as full-spectrum coverage for multiple disease applications. With the continuous development and inclusion of advanced algorithms, we expect this platform to largely simplify the clinical scientific research process and promote more and better discoveries.

Accurate delineation of individual teeth and alveolar bones from dental cone-beam CT images is an essential step in digital dentistry for precision dental healthcare. Here, the authors present a deep learning system for efficient, precise, and fully automatic segmentation of real-patient CBCT images presenting highly variable appearances.

Accurate brain tissue extraction on magnetic resonance imaging (MRI) data is crucial for analyzing brain structure and function. While several conventional tools have been optimized to handle human brain data, there have been no generalizable methods to extract brain tissues for multimodal MRI data from rodents, nonhuman primates, and humans. Therefore, developing a flexible and generalizable method for extracting whole brain tissue across species would allow researchers to analyze and compare experiment results more efficiently. Here, we propose a domain-adaptive and semi-supervised deep neural network, named the Brain Extraction Net (BEN), to extract brain tissues across species, MRI modalities, and MR scanners. We have evaluated BEN on 18 independent datasets, including 783 rodent MRI scans, 246 nonhuman primate MRI scans, and 4601 human MRI scans, covering five species, four modalities, and six MR scanners with various magnetic field strengths. Compared to conventional toolboxes, the superiority of BEN is illustrated by its robustness, accuracy, and generalizability. Our proposed method not only provides a generalized solution for extracting brain tissue across species but also significantly improves the accuracy of atlas registration, thereby benefiting the downstream processing tasks. As a novel fully automated deep-learning method, BEN is designed as an open-source software to enable high-throughput processing of neuroimaging data across species in preclinical and clinical applications.

Abstract Background The outbreak of coronavirus disease 2019 (COVID-19) causes tens of million infection world-wide. Many machine learning methods have been proposed for the computer-aided diagnosis between COVID-19 and community-acquired pneumonia (CAP) from chest computed tomography (CT) images. Most of these methods utilized the location-specific handcrafted features based on the segmentation results to improve the diagnose performance. However, the prerequisite segmentation step is time-consuming and needs the intervention by lots of expert radiologists, which cannot be achieved in the areas with limited medical resources. Methods We propose a generative adversarial feature completion and diagnosis network (GACDN) that simultaneously generates handcrafted features by radiomic counterparts and makes accurate diagnoses based on both original and generated features. Specifically, we first calculate the radiomic features from the CT images. Then, in order to fast obtain the location-specific handcrafted features, we use the proposed GACDN to generate them by its corresponding radiomic features. Finally, we use both radiomic features and location-specific handcrafted features for COVID-19 diagnosis. Results For the performance of our generated location-specific handcrafted features, the results of four basic classifiers show that it has an average of 3.21% increase in diagnoses accuracy. Besides, the experimental results on COVID-19 dataset show that our proposed method achieved superior performance in COVID-19 vs. community acquired pneumonia (CAP) classification compared with the state-of-the-art methods. Conclusions The proposed method significantly improves the diagnoses accuracy of COVID-19 vs. CAP in the condition of incomplete location-specific handcrafted features. Besides, it is also applicable in some regions lacking of expert radiologists and high-performance computing resources.

Abstract Background This study aimed to evaluate the utility of radiomics-based machine learning analysis with multiparametric DWI and to compare the diagnostic performance of radiomics features and mean diffusion metrics in the characterization of breast lesions. Methods This retrospective study included 542 lesions from February 2018 to November 2018. One hundred radiomics features were computed from mono-exponential (ME), biexponential (BE), stretched exponential (SE), and diffusion-kurtosis imaging (DKI). Radiomics-based analysis was performed by comparing four classifiers, including random forest (RF), principal component analysis (PCA), L1 regularization (L1R), and support vector machine (SVM). These four classifiers were trained on a training set with 271 patients via ten-fold cross-validation and tested on an independent testing set with 271 patients. The diagnostic performance of the mean diffusion metrics of ME (mADCall b, mADC0–1000), BE (mD, mD*, mf), SE (mDDC, mα), and DKI (mK, mD) were also calculated for comparison. The area under the receiver operating characteristic curve (AUC) was used to compare the diagnostic performance. Results RF attained higher AUCs than L1R, PCA and SVM. The AUCs of radiomics features for the differential diagnosis of breast lesions ranged from 0.80 (BE_D*) to 0.85 (BE_D). The AUCs of the mean diffusion metrics ranged from 0.54 (BE_mf) to 0.79 (ME_mADC0–1000). There were significant differences in the AUCs between the mean values of all diffusion metrics and radiomics features of AUCs (all P

Abstract Most prior studies focused on developing models for the severity or mortality prediction of COVID-19 patients. However, effective models for recovery-time prediction are still lacking. Here, we present a deep learning solution named iCOVID that can successfully predict the recovery-time of COVID-19 patients based on predefined treatment schemes and heterogeneous multimodal patient information collected within 48 hours after admission. Meanwhile, an interpretable mechanism termed FSR is integrated into iCOVID to reveal the features greatly affecting the prediction of each patient. Data from a total of 3008 patients were collected from three hospitals in Wuhan, China, for large-scale verification. The experiments demonstrate that iCOVID can achieve a time-dependent concordance index of 74.9% (95% CI: 73.6-76.3%) and an average day error of 4.4 days (95% CI: 4.2-4.6 days). Our study reveals that treatment schemes, age, symptoms, comorbidities, and biomarkers are highly related to recovery-time predictions.

As the leading cause of vision loss in working-age adults, diabetic retinopathy requires routinely retinal screening. Here the authors develop a deep learning system that can facilitate the screening by providing real-time image quality assessment, lesions detection, and grades across the disease spectrum.

Resting-state functional magnetic resonance imaging (rs-fMRI) is a non-invasive functional neuroimaging modality that has been widely used to investigate functional connectomes in the brain. Since noise and artifacts generated by non-neuronal physiological activities are predominant in raw rs-fMRI data, effective noise removal is one of the most important preprocessing steps prior to any subsequent analysis. For rs-fMRI denoising, a common trend is to decompose rs-fMRI data into multiple components and then regress out noise-related components. Therefore, various machine learning techniques have been used in such analyses with predefined procedures and manually engineered features. However, the lack of a universal definition of a noise-related source or artifact complicates manual feature engineering. Manual feature selection can result in the failure to capture unknown types of noise. Furthermore, the possibility that the hand-crafted features will only work for the broader population (e.g., healthy adults) but not for “outliers” (e.g., infants or subjects that belong to a disease cohort) is quite high. In practice, we have limited knowledge of which features should be extracted; thus, multi-classifier assembly must be implemented to improve performance, although this process is quite time-consuming. However, in real rs-fMRI applications, fast and accurate automatic identification of noise-related components on different datasets is critical. To solve this problem, we propose a novel, automatic, and end-to-end deep learning framework dedicated to noise-related component identification via a faster and more effective multi-layer feature extraction strategy that learns deeply embedded spatio-temporal features of the components. In this study, we achieved remarkable performance on various rs-fMRI datasets, including multiple adult rs-fMRI datasets from different rs-fMRI studies and an infant rs-fMRI dataset, which is quite heterogeneous and differs from that of adults. Our proposed framework also dramatically increases the noise detection speed owing to its inherent ability for deep learning (< 1s for single-component classification). It can be easily integrated into any preprocessing pipeline, even those that do not use standard procedures but depend on alternative toolboxes.

Summary: We described a challenge named “Diabetic Retinopathy (DR)—Grading and Image Quality Estimation Challenge” in conjunction with ISBI 2020 to hold three sub-challenges and develop deep learning models for DR image assessment and grading. The scientific community responded positively to the challenge, with 34 submissions from 574 registrations. In the challenge, we provided the DeepDRiD dataset containing 2,000 regular DR images (500 patients) and 256 ultra-widefield images (128 patients), both having DR quality and grading annotations. We discussed details of the top 3 algorithms in each sub-challenges. The weighted kappa for DR grading ranged from 0.93 to 0.82, and the accuracy for image quality evaluation ranged from 0.70 to 0.65. The results showed that image quality assessment can be used as a further target for exploration. We also have released the DeepDRiD dataset on GitHub to help develop automatic systems and improve human judgment in DR screening and diagnosis. The bigger picture: Diabetic retinopathy (DR) is the most common disease caused by diabetes. Challenges are held to address real-world issues encountered in the design of DR automated screening systems to advance the technology in this area. Thus, we described a challenge named ''Diabetic Retinopathy (DR)—Grading and Image Quality Estimation Challenge'' in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI 2020) for fundus image assessment and DR grading. The scientific community responded positively to the challenge. In the challenge, we provided a deep DR image dataset (DeepDRiD) containing regular DR images and ultra-widefield (UWF) DR images, both having image quality and DR grading diagnosis. We discussed details of the three best algorithms in each sub-challenges. The results by the top algorithms showed that image quality assessment can be used as a target for further exploration.

Abstract Background Spatial and temporal lung infection distributions of coronavirus disease 2019 (COVID-19) and their changes could reveal important patterns to better understand the disease and its time course. This paper presents a pipeline to analyze statistically these patterns by automatically segmenting the infection regions and registering them onto a common template. Methods A VB-Net is designed to automatically segment infection regions in CT images. After training and validating the model, we segmented all the CT images in the study. The segmentation results are then warped onto a pre-defined template CT image using deformable registration based on lung fields. Then, the spatial distributions of infection regions and those during the course of the disease are calculated at the voxel level. Visualization and quantitative comparison can be performed between different groups. We compared the distribution maps between COVID-19 and community acquired pneumonia (CAP), between severe and critical COVID-19, and across the time course of the disease. Results For the performance of infection segmentation, comparing the segmentation results with manually annotated ground-truth, the average Dice is 91.6% ± 10.0%, which is close to the inter-rater difference between two radiologists (the Dice is 96.1% ± 3.5%). The distribution map of infection regions shows that high probability regions are in the peripheral subpleural (up to 35.1% in probability). COVID-19 GGO lesions are more widely spread than consolidations, and the latter are located more peripherally. Onset images of severe COVID-19 (inpatients) show similar lesion distributions but with smaller areas of significant difference in the right lower lobe compared to critical COVID-19 (intensive care unit patients). About the disease course, critical COVID-19 patients showed four subsequent patterns (progression, absorption, enlargement, and further absorption) in our collected dataset, with remarkable concurrent HU patterns for GGO and consolidations. Conclusions By segmenting the infection regions with a VB-Net and registering all the CT images and the segmentation results onto a template, spatial distribution patterns of infections can be computed automatically. The algorithm provides an effective tool to visualize and quantify the spatial patterns of lung infection diseases and their changes during the disease course. Our results demonstrate different patterns between COVID-19 and CAP, between severe and critical COVID-19, as well as four subsequent disease course patterns of the severe COVID-19 patients studied, with remarkable concurrent HU patterns for GGO and consolidations.

The human brain is not only efficiently but also “redundantly” connected. The redundancy design could help the brain maintain resilience to disease attacks. This paper explores subnetwork-level redundancy dynamics and the potential of such metrics in disease studies. As such, we looked into specific functional subnetworks, including those associated with high-level functions. We investigated how the subnetwork redundancy dynamics change along with Alzheimer’s disease (AD) progression and with major depressive disorder (MDD), two major disorders that could share similar subnetwork alterations. We found an increased dynamic redundancy of the subcortical-cerebellum subnetwork and its connections to other high-order subnetworks in the mild cognitive impairment (MCI) and AD compared to the normal control (NC). With gained spatial specificity, we found such a redundancy index was sensitive to disease symptoms and could act as a protective mechanism to prevent the collapse of the brain network and functions. The dynamic redundancy of the medial frontal subnetwork and its connections to the frontoparietal subnetwork was also found decreased in MDD compared to NC. The spatial specificity of the redundancy dynamics changes may provide essential knowledge for a better understanding of shared neural substrates in AD and MDD.

Purpose: Parkinson’s disease (PD) diagnosis algorithms based on quantitative susceptibility mapping (QSM) and image algorithms rely on substantia nigra (SN) labeling. However, the difference between SN labels from different experts (or segmentation algorithms) will have a negative impact on downstream diagnostic tasks, such as the decrease of the accuracy of the algorithm or different diagnostic results for the same sample. In this article, we quantify the accuracy of the algorithm on different label sets and then improve the convolutional neural network (CNN) model to obtain a high-precision and highly robust diagnosis algorithm.Methods: The logistic regression model and CNN model were first compared for classification between PD patients and healthy controls (HC), given different sets of SN labeling. Then, based on the CNN model with better performance, we further proposed a novel “gated pooling” operation and integrated it with deep learning to attain a joint framework for image segmentation and classification.Results: The experimental results show that, with different sets of SN labeling that mimic different experts, the CNN model can maintain a stable classification accuracy at around 86.4%, while the conventional logistic regression model yields a large fluctuation ranging from 78.9 to 67.9%. Furthermore, the “gated pooling” operation, after being integrated for joint image segmentation and classification, can improve the diagnosis accuracy to 86.9% consistently, which is statistically better than the baseline.Conclusion: The CNN model, compared with the conventional logistic regression model using radiomics features, has better stability in PD diagnosis. Furthermore, the joint end-to-end CNN model is shown to be suitable for PD diagnosis from the perspectives of accuracy, stability, and convenience in actual use.

BackgroundEmerging evidence indicates that iron distribution is heterogeneous within the substantia nigra (SN) and it may reflect patient-specific trait of Parkinson’s Disease (PD). We assume it could account for variability in motor outcome of subthalamic nucleus deep brain stimulation (STN-DBS) in PD.ObjectiveTo investigate whether SN susceptibility features derived from radiomics with machine learning (RA-ML) can predict motor outcome of STN-DBS in PD.MethodsThirty-three PD patients underwent bilateral STN-DBS were recruited. The bilateral SN were segmented based on preoperative quantitative susceptibility mapping to extract susceptibility features using RA-ML. MDS-UPDRS III scores were recorded 1–3 days before and 6 months after STN-DBS surgery. Finally, we constructed three predictive models using logistic regression analyses: (1) the RA-ML model based on radiomics features, (2) the RA-ML+LCT (levodopa challenge test) response model which combined radiomics features with preoperative LCT response, (3) the LCT response model alone.ResultsFor the predictive performances of global motor outcome, the RA-ML model had 82% accuracy (AUC = 0.85), while the RA-ML+LCT response model had 74% accuracy (AUC = 0.83), and the LCT response model alone had 58% accuracy (AUC = 0.55). For the predictive performance of rigidity outcome, the accuracy of the RA-ML model was 80% (AUC = 0.85), superior to those of the RA-ML+LCT response model (76% accuracy, AUC = 0.82), and the LCT response model alone (58% accuracy, AUC = 0.42).ConclusionOur findings demonstrated that SN susceptibility features from radiomics could predict global motor and rigidity outcomes of STN-DBS in PD. This RA-ML predictive model might provide a novel approach to counsel candidates for STN-DBS.

Brain functional network (BFN) has become an increasingly important tool to explore individual differences and identify neurological/mental diseases. For estimating a “good” BFN (with more discriminative information for example), researchers have developed various methods, in which the most popular and simplest is Pearson's correlation (PC). Despite its empirical effectiveness, PC only encodes the low-order (second-order) statistics between brain regions. To model high-order statistics, researchers recently proposed to estimate BFN by conducting two sequential PCs (denoted as PC2 in this paper), and found that PC2-based BFN can provide additional information for group difference analysis. This inspires us to think about (1) what will happen if continuing the correlation operation to construct much higher-order BFN by PCn (n>2), and (2) whether the higher-order correlation will result in stronger discriminative ability. To answer these questions, we use PCn-based BFNs to predict individual differences (Female vs. Male) as well as identify subjects with mild cognitive impairment (MCI) from healthy controls (HCs). Through experiments, we have the following findings: (1) with the increase of n, the discriminative ability of PCn-based BFNs tends to decrease; (2) fusing the PCn-based BFNs (n>1) with the PC1-based BFN can generally improve the sensitivity for MCI identification, but fail to help the classification accuracy. In addition, we empirically find that the sequence of BFN adjacency matrices estimated by PCn (n = 1,2,3,⋯ ) will converge to a binary matrix with elements of ± 1.

The ongoing worldwide COVID-19 pandemic has become a huge threat to global public health. Using CT image, 3389 COVID-19 patients, 1593 community-acquired pneumonia (CAP) patients, and 1707 nonpneumonia subjects were included to explore the different patterns of lung and lung infection. We found that COVID-19 patients have a significant reduced lung volume with increased density and mass, and the infections tend to present as bilateral lower lobes. The findings provide imaging evidence to improve our understanding of COVID-19.

The sliding-window-based dynamic functional connectivity network (D-FCN) has been becoming an increasingly useful tool for understanding the changes of brain connectivity patterns and the association of neurological diseases with these dynamic variations. However, conventional D-FCN is essentially low-order network, which only reflects the pairwise interaction pattern between brain regions and thus overlooking the high-order interactions among multiple brain regions. In addition, D-FCN is innate with temporal sensitivity issue, i.e., D-FCN is sensitive to the chronological order of its subnetworks. To deal with the above issues, we propose a novel high-order functional connectivity network framework based on the central moment feature of D-FCN. Specifically, we firstly adopt a central moment approach to extract multiple central moment feature matrices from D-FCN. Furthermore, we regard the matrices as the profiles to build multiple high-order functional connectivity networks which further capture the higher level and more complex interaction relationships among multiple brain regions. Finally, we use the voting strategy to combine the high-order networks with D-FCN for autism spectrum disorder diagnosis. Experimental results show that the combination of multiple functional connectivity networks achieves accuracy of 88.06%, and the best single network achieves accuracy of 79.5%.

Perivascular spaces (PVS) in the human brain are related to various brain diseases. However, it is difficult to quantify them due to their thin and blurry appearance. In this paper, we introduce a deep-learning-based method, which can enhance a magnetic resonance (MR) image to better visualize the PVS. To accurately predict the enhanced image, we propose a very deep 3D convolutional neural network that contains densely connected networks with skip connections. The proposed networks can utilize rich contextual information derived from low-level to high-level features and effectively alleviate the gradient vanishing problem caused by the deep layers. The proposed method is evaluated on 17 7T MR images by a twofold cross-validation. The experiments show that our proposed network is much more effective to enhance the PVS than the previous PVS enhancement methods.

Hippocampal segmentation from infant brain MR images is indispensable for studying early brain development. However, most of the hippocampal segmentation methods were developed for population-based adult brain images, which are not suitable for longitudinal infant brain images acquired in the first year of life due to the low image contrast and variable development patterns of the hippocampal structure. To address these challenges, we propose a classification-guided boundary regression method to first detect hippocampal boundaries in the longitudinal infant brain images and then use those detected boundaries to guide the deformable model for final segmentation. Specifically, we first employ a classification-guided regression forest to predict the 3D displacements from individual image voxels to the potential hippocampal boundaries. These predicted displacements then determine the boundary maps by a voting strategy. Second, we iteratively enhance the voted hippocampal boundary map by incorporating the spatial context information given the tentative boundary estimation of the current time point. Besides, the longitudinal context information from all time points of the temporal sequence of the same subject (i.e., given their tentative segmentation results) is also utilized to facilitate accurate segmentation. Finally, a deformable model is applied to the enhanced voted boundary maps for achieving the longitudinal hippocampal segmentation. The experiments on infant brain MR images acquired from 2-week-old to 1-year-old show promising hippocampal segmentation results, indicating the applicability of our method in early brain development study.

Multimodal heterogeneous data, such as structural magnetic resonance imaging (MRI), positron emission tomography (PET), and cerebrospinal fluid (CSF), are effective in improving the performance of automated dementia diagnosis by providing complementary information on degenerated brain disorders, such as Alzheimer's prodromal stage, i.e., mild cognitive impairment. Effectively integrating multimodal data has remained a challenging problem, especially when these heterogeneous data are incomplete due to poor data quality and patient dropout. Besides, multimodal data usually contain noise information caused by different scanners or imaging protocols. The existing methods usually fail to well handle these heterogeneous and noisy multimodal data for automated brain dementia diagnosis. To this end, we propose a high-order Laplacian regularized low-rank representation method for dementia diagnosis using block-wise missing multimodal data. The proposed method was evaluated on 805 subjects (with incomplete MRI, PET, and CSF data) from the real Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort. Experimental results suggest the effectiveness of our method in three tasks of brain disease classification, compared with the state-of-the-art methods.

The sliding-window-based dynamic functional connectivity networks (SW-D-FCN) derive from resting-state functional Magnetic Resonance Imaging has become an increasingly useful tool in the diagnosis of various neurodegenerative diseases. However, it is still challenging to learn how to extract and select the most discriminative features from SW-D-FCN. Conventionally, existing methods opt to select a single discriminative feature set or concatenate a few more from the SW-D-FCN. However, such reductionist strategies may fail to fully capture the personalized discriminative characteristics contained in each functional connectivity (FC) sequence of the SW-D-FCN. To address this issue, we propose a unit-based personalized fingerprint feature selection (UPFFS) strategy to better capture the most discriminative feature associated with a target disease for each unit. Specifically, we regard the FC sequence between any pair of brain regions of interest (ROIs) is regarded as a unit. For each unit, the most discriminative feature is identified by a specific feature evaluation method and all the most discriminative features are then concatenated together as a feature set for the subsequent classification task. In such a way, the personalized fingerprint feature derived from each FC sequence can be fully mined and utilized in classification decision. To illustrate the effectiveness of the proposed strategy, we conduct experiments to distinguish subjects diagnosed with autism spectrum disorder from normal controls. Experimental results show that the proposed strategy can select relevant discriminative features and achieve superior performance to benchmark methods.

As non-human primates, macaques have a close phylogenetic relationship to human beings and have been proven to be a valuable and widely used animal model in human neuroscience research. Accurate skull stripping (aka. brain extraction) of brain magnetic resonance imaging (MRI) is a crucial prerequisite in neuroimaging analysis of macaques. Most of the current skull stripping methods can achieve satisfactory results for human brains, but when applied to macaque brains, especially during early brain development, the results are often unsatisfactory. In fact, the early dynamic, regionally-heterogeneous development of macaque brains, accompanied by poor and age-related contrast between different anatomical structures, poses significant challenges for accurate skull stripping. To overcome these challenges, we propose a fully-automated framework to effectively fuse the age-specific intensity information and domain-invariant prior knowledge as important guiding information for robust skull stripping of developing macaques from 0 to 36 months of age. Specifically, we generate Signed Distance Map (SDM) and Center of Gravity Distance Map (CGDM) based on the intermediate segmentation results as guidance. Instead of using local convolution, we fuse all information using the Dual Self-Attention Module (DSAM), which can capture global spatial and channel-dependent information of feature maps. To extensively evaluate the performance, we adopt two relatively-large challenging MRI datasets from rhesus macaques and cynomolgus macaques, respectively, with a total of 361 scans from two different scanners with different imaging protocols. We perform cross-validation by using one dataset for training and the other one for testing. Our method outperforms five popular brain extraction tools and three deep-learning-based methods on cross-source MRI datasets without any transfer learning.

Brain functional network (BFN) analysis is becoming a crucial way to explore the inherent organized pattern of the brain and reveal potential biomarkers for diagnosing neurological or psychological disorders. In so doing, a well-estimated BFN is of great concern. In practice, however, noises or artifacts involved in the observed data (i.e., fMRI time series in this paper) generally lead to a poor estimation of BFN, and thus a complex preprocessing pipeline is often used to improve the quality of the data prior to BFN estimation. One of the popular preprocessing steps is data-scrubbing that aims at removing “bad” volumes from the fMRI time series according to the amplitude of the head motion. Despite its helpfulness in general, this traditional scrubbing scheme cannot guarantee that the removed volumes are necessarily unhelpful, since such a step is fully independent to the subsequent BFN estimation task. Moreover, the removal of volumes would reduce the statistical power, and different numbers of volumes are generally scrubbed for different subjects, resulting in an inconsistency or bias in the estimated BFNs. To address these issues, we develop a new learning framework that conducts BFN estimation and data-scrubbing simultaneously by an alternating optimization algorithm. The newly developed algorithm adaptively weights volumes (instead of removing them directly) for the task of BFN estimation. As a result, the proposed method can not only reduce the difficulty of threshold selection involved in the traditional scrubbing scheme, but also provide a more flexible framework that scrubs the data in the subsequent FBN estimation model. Finally, we validate the proposed method by identifying subjects with mild cognitive impairment (MCI) from normal controls based on the estimated BFNs, achieving an 80.22% classification accuracy, which significantly improves the baseline methods.

Major depressive disorder (MDD) represents a grand challenge to human health and society, but the underlying pathophysiological mechanisms remain elusive. Previous neuroimaging studies have suggested that MDD is associated with abnormal interactions and dynamics in two major neural systems including the default mode - salience (DMN-SAL) network and the executive - limbic (EXE-LIM) network, but it is not clear which network plays a central role and which network plays a subordinate role in MDD pathophysiology. To address this question, we refined a newly developed Multiscale Neural Model Inversion (MNMI) framework and applied it to test whether MDD is more affected by impaired circuit interactions in the DMN-SAL network or the EXE-LIM network. The model estimates the directed connection strengths between different neural populations both within and between brain regions based on resting-state fMRI data collected from normal healthy subjects and patients with MDD. Results show that MDD is primarily characterized by abnormal circuit interactions in the EXE-LIM network rather than the DMN-SAL network. Specifically, we observe reduced frontoparietal effective connectivity that potentially contributes to hypoactivity in the dorsolateral prefrontal cortex (dlPFC), and decreased intrinsic inhibition combined with increased excitation from the superior parietal cortex (SPC) that potentially lead to amygdala hyperactivity, together resulting in activation imbalance in the PFC-amygdala circuit that pervades in MDD. Moreover, the model reveals reduced PFC-to-hippocampus excitation but decreased SPC-to-thalamus inhibition in MDD population that potentially lead to hypoactivity in the hippocampus and hyperactivity in the thalamus, consistent with previous experimental data. Overall, our findings provide strong support for the long-standing limbic-cortical dysregulation model in major depression but also offer novel insights into the multiscale pathophysiology of this debilitating disease.

[This corrects the article DOI: 10.1371/journal.pone.0235039.].

Accurate removal of magnetic resonance imaging (MRI) signal outside the brain, a.k.a., skull stripping, is a key step in the brain image pre-processing pipelines. In rodents, this is mostly achieved by manually editing a brain mask, which is time-consuming and operator dependent. Automating this step is particularly challenging in rodents as compared to humans, because of differences in brain/scalp tissue geometry, image resolution with respect to brain-scalp distance, and tissue contrast around the skull. In this study, we proposed a deep-learning-based framework, U-Net, to automatically identify the rodent brain boundaries in MR images. The U-Net method is robust against inter-subject variability and eliminates operator dependence. To benchmark the efficiency of this method, we trained and validated our model using both in-house collected and publicly available datasets. In comparison to current state-of-the-art methods, our approach achieved superior averaged Dice similarity coefficient to ground truth T2-weighted rapid acquisition with relaxation enhancement and T2∗-weighted echo planar imaging data in both rats and mice (all p < 0.05), demonstrating robust performance of our approach across various MRI protocols.

Perivascular spaces (PVSs) are fluid-filled spaces surrounding penetrating blood vessels in the brain and are an integral pathway of the glymphatic system. A PVS and the enclosed blood vessel are commonly visualized as a single vessel-like complex (denoted as PVSV) in high-resolution MRI images. Quantitative characterization of the PVSV morphology in MRI images in healthy subjects may serve as a reference for detecting disease related PVS and/or blood vessel alterations in patients with brain diseases.To this end, we evaluated the age dependences, spatial heterogeneities, and dynamic properties of PVSV morphological features in 45 healthy subjects (21–55 years old), using an ultra-high-resolution three-dimensional transverse relaxation time weighted MRI sequence (0.41 ​× ​0.41 ​× ​0.4 ​mm3) at 7T. Quantitative PVSV parameters, including apparent diameter, count, volume fraction (VF), and relative contrast to noise ratio (rCNR) were calculated in the white matter and subcortical structures. Dynamic changes were induced by carbogen breathing which are known to induce vasodilation and increase the blood oxygenation level in the brain.PVSV count and VF significantly increased with age in basal ganglia (BG), so did rCNR in BG, midbrain, and white matter (WM). Apparent PVSV diameter also showed a positive association with age in the three brain regions, although it did not reach statistical significance. The PVSV VF and count showed large inter-subject variations, with coefficients of variation ranging from 0.17 to 0.74 after regressing out age and gender effects. Both apparent diameter and VF exhibited significant spatial heterogeneity, which cannot be explained solely by radio-frequency field inhomogeneities. Carbogen breathing significantly increased VF in BG and WM, and rCNR in thalamus, BG, and WM compared to air breathing.Our results are consistent with gradual dilation of PVSs with age in healthy adults. The PVSV morphology exhibited spatial heterogeneity and large inter-subject variations and changed during carbogen breathing compared to air breathing.

The sliding-window-based dynamic functional connectivity networks (D-FCNs) derived from resting-state functional magnetic resonance imaging (rs-fMRI) are effective methods for diagnosing various neurological diseases, including autism spectrum disorder (ASD). However, traditional D-FCNs are low-order networks based on pairwise correlation between brain regions, thus overlooking high-level interactions across multiple regions of interest (ROIs). Moreover, D-FCNs suffer from the temporal mismatching issue, i.e., subnetworks in the same temporal window do not have temporal correspondence across different subjects. To address the above problems, we first construct a novel high-order D-FCNs based on the principle of “correlation’s correlation” to further explore the higher level and more complex interaction relationships among multiple ROIs. Furthermore, we propose to use a central-moment method to extract temporal-invariance properties contained in either low- or high-order D-FCNs. Finally, we design and train an ensemble classifier by fusing the features extracted from conventional FCN, low-order D-FCNs, and high-order D-FCNs for the diagnosis of ASD and normal control subjects. Our method achieved the best ASD classification accuracy (83%), and our results revealed the features extracted from different networks fingerprinting the autistic brain at different connectional levels.

MR Fingerprinting (MRF) is a relatively new imaging framework capable of providing accurate and simultaneous quantification of multiple tissue properties for improved tissue characterization and disease diagnosis. While 2D MRF has been widely available, extending the method to 3D MRF has been an actively pursued area of research as a 3D approach can provide a higher spatial resolution and better tissue characterization with an inherently higher signal-to-noise ratio. However, 3D MRF with a high spatial resolution requires lengthy acquisition times, especially for a large volume, making it impractical for most clinical applications. In this study, a high-resolution 3D MR Fingerprinting technique, combining parallel imaging and deep learning, was developed for rapid and simultaneous quantification of T1 and T2 relaxation times. Parallel imaging was first applied along the partition-encoding direction to reduce the amount of acquired data. An advanced convolutional neural network was then integrated with the MRF framework to extract features from the MRF signal evolution for improved tissue characterization and accelerated mapping. A modified 3D-MRF sequence was also developed in the study to acquire data to train the deep learning model that can be directly applied to prospectively accelerate 3D MRF scans. Our results of quantitative T1 and T2 maps demonstrate that improved tissue characterization can be achieved using the proposed method as compared to prior methods. With the integration of parallel imaging and deep learning techniques, whole-brain (26 × 26 × 18 cm3) quantitative T1 and T2 mapping with 1-mm isotropic resolution were achieved in ~7 min. In addition, a ~7-fold improvement in processing time to extract tissue properties was also accomplished with the deep learning approach as compared to the standard template matching method. All of these improvements make high-resolution whole-brain quantitative MR imaging feasible for clinical applications.

Biomedical imaging Is an important source of information in cancer research. Characterizations of cancer morphology at onset, progression, and in response to treatment provide complementary information to that gleaned from genomics and clinical data. Accurate extraction and classification of both visual and latent image features Is an increasingly complex challenge due to the increased complexity and resolution of biomedical image data. In this paper, we present four deep learning-based image analysis methods from the Computational Precision Medicine (CPM) satellite event of the 21st International Medical Image Computing and Computer Assisted Intervention (MICCAI 2018) conference. One method Is a segmentation method designed to segment nuclei in whole slide tissue images (WSIs) of adult diffuse glioma cases. It achieved a Dice similarity coefficient of 0.868 with the CPM challenge datasets. Three methods are classification methods developed to categorize adult diffuse glioma cases into oligodendroglioma and astrocytoma classes using radiographic and histologic image data. These methods achieved accuracy values of 0.75, 0.80, and 0.90, measured as the ratio of the number of correct classifications to the number of total cases, with the challenge datasets. The evaluations of the four methods indicate that (1) carefully constructed deep learning algorithms are able to produce high accuracy in the analysis of biomedical image data and (2) the combination of radiographic with histologic image information improves classification performance.

Functional brain network (FBN), estimated with functional magnetic resonance imaging (fMRI), has become a potentially useful way of diagnosing neurological disorders in their early stages by comparing the connectivity patterns between different brain regions across subjects. However, this depends, to a great extent, on the quality of the estimated FBNs, indicating that FBN estimation is a key step for the subsequent task of disorder identification. In the past decades, researchers have developed many methods to estimate FBNs, including Pearson's correlation and (regularized) partial correlation, etc. Despite their widespread applications in current studies, most of the existing methods estimate FBNs only based on the dependency between the measured blood oxygen level dependent (BOLD) signals, which ignores spatial relationship of signals associated with different brain regions. Due to the space and material parsimony principle of our brain, we believe that the spatial distance between brain regions has an important influence on FBN topology. Therefore, in this paper, we assume that spatially neighboring brain regions tend to have stronger connections and/or share similar connections with others; based on this assumption, we propose two novel methods to estimate FBNs by incorporating the information of brain region distance into the estimation model. To validate the effectiveness of the proposed methods, we use the estimated FBNs to identify subjects with mild cognitive impairment (MCI) from normal controls (NCs). Experimental results show that the proposed methods are better than the baseline methods in the sense of MCI identification accuracy.

Background: Recent evidence suggests the presence of hippocampal neuroanatomical abnormalities in subjects of amnestic mild cognitive impairment (aMCI). Our study aimed to identify the radiomic biomarkers of the hippocampus for building the classification models in aMCI diagnosis.Methods: For this target, we recruited 42 subjects with aMCI and 44 normal controls (NC). The right and left hippocampi were segmented for each subject using an efficient learning-based method. Then, the radiomic analysis was applied to calculate and select the radiomic features. Finally, two logistic regression models were built based on the selected features obtained from the right and left hippocampi.Results: There were 385 features derived after calculation, and four features remained after feature selection from each group of data. The area under the receiver operating characteristic (ROC) curve, specificity, sensitivity, positive predictive value, negative predictive value, precision, recall, and F-score of the classification evaluation index of the right hippocampus logistic regression model were 0.76, 0.71, 0.69, 0.69, 0.71, 0.69, 0.69, and 0.69, and those of the left hippocampus model were 0.79, 0.71, 0.54, 0.64, 0.63, 0.64, 0.54, and 0.58, respectively.Conclusion: Results demonstrate the potential hippocampal radiomic biomarkers are valid for the aMCI diagnosis. The MRI-based radiomic analysis, with further improvement and validation, can be used to identify patients with aMCI and guide the individual treatment.

Abstract BM3D is a state-of-the-art image denoising method. Its denoised results in the regions with strong edges can often be better than in the regions with smooth or weak edges, due to more accurate block-matching for the strong-edge regions. So using adaptive block sizes on different image regions may result in better image denoising. Based on these observations, in this paper, we first partition each image into regions belonging to one of the three morphological components, i.e., contour, texture, and smooth components, according to the regional energy of alternating current (AC) coefficients of discrete cosine transform (DCT). Then, we can adaptively determine the block size for each morphological component. Specifically, we use the smallest block size for the contour components, the medium block size for the texture components, and the largest block size for the smooth components. To better preserve image details, we also use a multi-stage strategy to implement image denoising, where every stage is similar to the BM3D method, except using adaptive sizes and different transform dimensions. Experimental results show that our proposed algorithm can achieve higher PSNR and MSSIM values than the BM3D method, and also better visual quality of denoised images than by the BM3D method and some other existing state-of-the-art methods.

Retinal layer segmentation from optical coherence tomography (OCT) is of fundamental importance for measuring retinal layer thicknesses. These thickness measurements have been shown to correlate well with the severity of different ocular diseases; hence, they provide useful diagnostic information concerning diseases. Manual segmentation of retinal layers from OCT remains dominant in ophthalmological clinical practice but has serious drawbacks: it is time consuming, labor intensive, and results in inter/intrarater variations. Computer aided segmentation has attracted intensive research attention because it holds the potential not only to provide repeatable, quantitative, and objective results but also to reduce the time and effort required to delineate the retinal layers. However, most of the existing computer based retinal layer segmentation techniques focus on segmenting specific layers by exploring their unique characteristics; thus, they can fail to segment a retinal layer that is totally different. In this paper, we propose a generative retinal layer segmentation method based on groupwise curve alignment that combines the capabilities of segmenting different retinal layers into a unified framework. This method is unique for both its accuracy and its ability to segment any retinal layer without any special modifications. We experimentally validate that the proposed method outperforms a representative state-of-the-art technique by using images of both normal healthy eyes and diseased eyes. Our method is potentially useful in a large variety of practical applications involving retinal layer segmentation from OCT.

Abstract Perivascular spaces (PVSs) in brain have a close relationship with typical neurological diseases. The quantitative studies of PVSs are meaningful but usually difficult, due to their thin and weak signals and also background noise in the 7 T brain magnetic resonance images (MRI). To clearly distinguish the PVSs in the 7 T MRI, we propose a novel PVS enhancement method based on the Haar transform of non-local cubes. Specifically, we extract a certain number of cubes from a small neighbor to form a cube group, and then perform Haar transform on each cube group. The Haar transform coefficients are processed using a nonlinear function to amplify the weak signals relevant to the PVSs and to suppress the noise. The enhanced image is reconstructed using the inverse Haar transform of the processed coefficients. Finally, we perform a block-matching 4D filtering on the enhanced image to further remove any remaining noise, and thus obtain an enhanced and denoised 7 T MRI for PVS segmentation. We apply two existing methods to complete PVS segmentation, i.e., (1) vesselness-thresholding and (2) random forest classification. The experimental results show that the PVS segmentation performances can be significantly improved by using the enhanced and denoised 7 T MRI.

Abstract Conventional functional connectivity (FC), referred to as low-order FC, estimates temporal correlation of the resting-state functional magnetic resonance imaging (rs-fMRI) time series between any pair of brain regions, simply ignoring the potentially high-level relationship among these brain regions. A high-order FC based on “correlation’s correlation” has emerged as a new approach for abnormality detection of brain disease. However, separate construction of the low- and high-order FC networks overlooks information exchange between the two FC levels. Such a higher-level relationship could be more important for brain diseases study. In this paper, we propose a novel framework, namely “hybrid high-order FC networks” by exploiting the higher-level dynamic interaction among brain regions for early mild cognitive impairment (eMCI) diagnosis. For each sliding window-based rs-fMRI sub-series, we construct a whole-brain associated high-order network, by estimating the correlations between the topographical information of the high-order FC sub-network from one brain region and that of the low-order FC sub-network from another brain region. With multi-kernel learning, complementary features from multiple time-varying FC networks constructed at different levels are fused for eMCI classification. Compared with other state-of-the-art methods, the proposed framework achieves superior diagnosis accuracy, and hence could be promising for understanding pathological changes of brain connectome.

Groupwise image registration tackles biases that can potentially arise from inappropriate template selection. It typically involves simultaneous registration of a cohort of images to a common space that is not specified a priori. Existing groupwise registration methods are computationally complex and are only effective for image populations without large anatomical variations. In this paper, we propose a deep learning framework to rapidly estimate large deformations between images to significantly reduce structural variability. Specifically, we employ a multi-level graph coarsening method to agglomerate similar images into clusters, each represented by an exemplar image. We then use a deep learning framework to predict the initial deformations between images. Warping with the estimated deformations brings the images closer in the image manifold and their alignment can be further refined using conventional groupwise registration algorithms. We evaluated the effectiveness of our method in groupwise registration of MR brain images and compared it against state-of-the-art groupwise registration methods. Experimental results indicate that deformation initialization enables groupwise registration to converge significantly faster with competitive accuracy, therefore facilitates large-scale imaging studies.

Accurate and automatic segmentation of infant hippocampal subfields from magnetic resonance (MR) images is an important step for studying memory related infant neurological diseases. However, existing hippocampal subfield segmentation methods were generally designed based on adult subjects, and would compromise performance when applied to infant subjects due to insufficient tissue contrast and fast changing structural patterns of early hippocampal development. In this paper, we propose a new fully convolutional network (FCN) for infant hippocampal subfield segmentation by embedding the dilated dense network in the U-net, namely DUnet. The embedded dilated dense network can generate multi-scale features while keeping high spatial resolution, which is useful in fusing the low-level features in the contracting path with the high-level features in the expanding path. To further improve the performance, we group every pair of convolutional layers with one residual connection in the DUnet, and obtain the Residual DUnet (ResDUnet). Experimental results show that our proposed DUnet and ResDUnet improve the average Dice coefficient by 2.1 and 2.5% for infant hippocampal subfield segmentation, respectively, when compared with the classic 3D U-net. The results also demonstrate that our methods outperform other state-of-the-art methods.

Analysis of developmental brain networks is fundamentally important for basic developmental neuroscience. In this paper, we focus on the temporally-covarying connection patterns, called meta-networks, and develop a new mathematical model for meta-network decomposition. With the proposed model, we decompose the developmental structural correlation networks of cortical thickness into five meta-networks. Each meta-network exhibits a distinctive spatial connection pattern, and its covarying trajectory highlights the temporal contribution of the meta-network along development. Systematic analysis of the meta-networks and covarying trajectories provides insights into three important aspects of brain network development.

Objectives: We aimed to investigate whether an inter-voxel diffusivity metric (local diffusion homogeneity, LDH), can provide supplementary information to traditional intra-voxel metrics (i.e., fractional anisotropy, FA) in white matter (WM) abnormality detection for type 2 diabetes mellitus (T2DM).Methods: Diffusion tensor imaging was acquired from 34 T2DM patients and 32 healthy controls. Voxel-based group-difference comparisons based on LDH and FA, as well as the association between the diffusion metrics and T2DM risk factors [i.e., body mass index (BMI) and systolic blood pressure (SBP)], were conducted, with age, gender and education level controlled.Results: Compared to the controls, T2DM patients had higher LDH in the pons and left temporal pole, as well as lower FA in the left superior corona radiation (p < 0.05, corrected). In T2DM, there were several overlapping WM areas associated with BMI as revealed by both LDH and FA, including right temporal lobe and left inferior parietal lobe; but the unique areas revealed only by using LDH included left inferior temporal lobe, right supramarginal gyrus, left pre- and post-central gyrus (at the semiovale center), and right superior radiation. Overlapping WM areas that associated with SBP were found with both LDH and FA, including right temporal pole, bilateral orbitofrontal area (rectus gyrus), the media cingulum bundle, and the right cerebellum crus I. However, the unique areas revealed only by LDH included right inferior temporal lobe, right inferior occipital lobe, and splenium of corpus callosum.Conclusion: Inter- and intra-voxel diffusivity metrics may have different sensitivity in the detection of T2DM-related WM abnormality. We suggested that LDH could provide supplementary information and reveal additional underlying brain changes due to diabetes.