Search

The research groups in computer vision, graphics, and machine learning have dedicated a substantial amount of attention to the areas of 3D object reconstruction, augmentation, and registration. Deep learning is the predominant method used in artificial intelligence for addressing computer vision challenges. However, deep learning on three-dimensional data presents distinct obstacles and is now in its nascent phase. There have been significant advancements in deep learning specifically for three-dimensional data, offering a range of ways to address these issues. This study offers a comprehensive examination of the latest advancements in deep learning methodologies. We examine many benchmark models for the tasks of 3D object registration, augmentation, and reconstruction. We thoroughly analyse their architectures, advantages, and constraints. In summary, this report provides a comprehensive overview of recent advancements in three-dimensional deep learning and highlights unresolved research areas that will need to be addressed in the future.

Quality assessment in industrial applications is often carried out through visual inspection, usually performed or supported by human domain experts. However, the manual visual inspection of processes and products is error-prone and expensive. It is therefore not surprising that the automation of visual inspection in manufacturing and maintenance is heavily researched and discussed. The use of artificial intelligence as an approach to visual inspection in industrial applications has been considered for decades. Recent successes, driven by advances in deep learning, present a possible paradigm shift and have the potential to facilitate automated visual inspection, even under complex environmental conditions. For this reason, we explore the question of to what extent deep learning is already being used in the field of automated visual inspection and which potential improvements to the state of the art could be realized utilizing concepts from academic research. By conducting an extensive review of the openly accessible literature, we provide an overview of proposed and in-use deep-learning models presented in recent years. Our survey consists of 196 open-access publications, of which 31.7% are manufacturing use cases and 68.3% are maintenance use cases. Furthermore, the survey also shows that the majority of the models currently in use are based on convolutional neural networks, the current de facto standard for image classification, object recognition, or object segmentation tasks. Nevertheless, we see the emergence of vision transformer models that seem to outperform convolutional neural networks but require more resources, which also opens up new research opportunities for the future. Another finding is that in 97% of the publications, the authors use supervised learning techniques to train their models. However, with the median dataset size consisting of 2500 samples, deep-learning models cannot be trained from scratch, so it would be beneficial to use other training paradigms, such as self-supervised learning. In addition, we identified a gap of approximately three years between approaches from deep-learning-based computer vision being published and their introduction in industrial visual inspection applications. Based on our findings, we additionally discuss potential future developments in the area of automated visual inspection.

Rock Paper Scissors is one of the most popular games in the world, because of their easy and simple way to play among young and elderly people. The point of this game is to do the draw or just to find out who loses or wins. The pandemic conditions made people unable to meet face-to-face and could only play this game virtually. To carry out this activity in a virtual way, this research facilitates a model in the form of image classification to distinguish the hand gestures s in the form of rock, paper, and scissors. This classification process utilizes the Convolutional Neural Network (CNN) method. This method is one type of artificial neural network in terms of image classification. CNN uses three stages, namely convolutional layer, pooling layer, and fully connected layer. The implementation of this method for hand gesture classification in the form of rock, scissors, and paper images in this study shows an increased average accuracy towards the previous study from 97.66% to 99%.

This document is a summary of a special issue of the journal Sensors, titled "Emerging AI+X-Based Sensor and Networking Technologies including Selected Papers from ICGHIT 2022–2023." The special issue features selected papers from the 10th and 11th International Conferences on Green and Human Information Technology (ICGHITs), which were held in Korea and Thailand. The conferences focused on the theme of "Emerging Artificial Intelligent (AI)+X technology" and "Hyper Automation + Human AI" respectively. The selected papers cover various topics such as network security, routing protocols, signal detection, and clustering mechanisms, all incorporating AI-based methods. The issue also includes papers on topics like secure authentication, distance estimation in RFID systems, energy optimization in smart homes, blockchain technology, and radar signal detection. The authors emphasize the importance of both technology and humanity in advancing green and information technologies. [Extracted from the article]

Given the recent events worldwide due to viral diseases that affect human health, automatic monitoring systems are one of the strong points of research that has gained strength, where the detection of biohazardous waste of a sanitary nature is highlighted related to viral diseases stands out. It is essential in this field to generate developments aimed at saving lives, where robotic systems can operate as assistants in various fields. In this work an artificial intelligence algorithm based on two stages is presented, one is the recognition of paper debris using a ResNet-50, chosen for its object localization capacity, and the other is a fuzzy inference system for the generation of alarm states due to biological risk by such debris, where fuzzy logic helps to establish a model for a non-predictive system as the one exposed. A biohazard detection algorithm for paper waste is described, oriented to operate on an assistive robot in a residential environment. The training parameters of the network, which achieve 100% accuracy with confidence levels between 82% for very small waste and 100% in direct view, are presented. Timing cycles are established for validation of the exposure time of the waste, where through the fuzzy system, risk alarms are generated, which allows establishing a system with an average reliability of 98%. [ABSTRACT FROM AUTHOR]

Predictive maintenance strategies are becoming increasingly more important with the increased needs for automation and digitalization within pulp and paper manufacturing sector. Hence, this study contributes to examine the most efficient pre-processing approaches for predicting sensory data trends based on Gated Recurrent Unit (GRU) neural networks. To validate the model, the data from two paper pulp presses with several pre-processing methods are utilized for predicting the units' conditions. The results of validation criteria show that pre-processing data using a LOWESS in combination with the Elimination of discrepant data filter achieves more stable results, the prediction error decreases, and the predicted values are easier to interpret. The model can anticipate future values with MAPE, RMSE and MAE of 1.2, 0.27 and 0.30 respectively. The errors are below the significance level. Moreover, it is identified that the best hyperparameters found for each paper pulp press must be different. [ABSTRACT FROM AUTHOR]

Background: Preclinical low-count positron emission tomography (LC-PET) imaging offers numerous advantages such as facilitating imaging logistics, enabling longitudinal studies of long- and short-lived isotopes as well as increasing scanner throughput. However, LC-PET is characterized by reduced photon-count levels resulting in low signal-to-noise ratio (SNR), segmentation difficulties, and quantification uncertainties., Purpose: We developed and evaluated a novel deep-learning (DL) architecture-Attention based Residual-Dilated Net (ARD-Net)-to generate standard-count PET (SC-PET) images from LC-PET images. The performance of the ARD-Net framework was evaluated for numerous low count realizations using fidelity-based qualitative metrics, task-based segmentation, and quantitative metrics., Method: Patient Derived tumor Xenograft (PDX) with tumors implanted in the mammary fat-pad were subjected to preclinical [ 18 F]-Fluorodeoxyglucose (FDG)-PET/CT imaging. SC-PET images were derived from a 10 min static FDG-PET acquisition, 50 min post administration of FDG, and were resampled to generate four distinct LC-PET realizations corresponding to 10%, 5%, 1.6%, and 0.8% of SC-PET count-level. ARD-Net was trained and optimized using 48 preclinical FDG-PET datasets, while 16 datasets were utilized to assess performance. Further, the performance of ARD-Net was benchmarked against two leading DL-based methods (Residual UNet, RU-Net; and Dilated Network, D-Net) and non-DL methods (Non-Local Means, NLM; and Block Matching 3D Filtering, BM3D). The performance of the framework was evaluated using traditional fidelity-based image quality metrics such as Structural Similarity Index Metric (SSIM) and Normalized Root Mean Square Error (NRMSE), as well as human observer-based tumor segmentation performance (Dice Score and volume bias) and quantitative analysis of Standardized Uptake Value (SUV) measurements. Additionally, radiomics-derived features were utilized as a measure of quality assurance (QA) in comparison to true SC-PET. Finally, a performance ensemble score (EPS) was developed by integrating fidelity-based and task-based metrics. Concordance Correlation Coefficient (CCC) was utilized to determine concordance between measures. The non-parametric Friedman Test with Bonferroni correction was used to compare the performance of ARD-Net against benchmarked methods with significance at adjusted p-value ≤0.01., Results: ARD-Net-generated SC-PET images exhibited significantly better (p ≤ 0.01 post Bonferroni correction) overall image fidelity scores in terms of SSIM and NRMSE at majority of photon-count levels compared to benchmarked DL and non-DL methods. In terms of task-based quantitative accuracy evaluated by SUV Mean  and SUV Peak, ARD-Net exhibited less than 5% median absolute bias for SUV Mean compared to true SC-PET and lower degree of variability compared to benchmarked DL and non-DL based methods in generating SC-PET. Additionally, ARD-Net-generated SC-PET images displayed higher degree of concordance to SC-PET images in terms of radiomics features compared to non-DL and other DL approaches. Finally, the ensemble score suggested that ARD-Net exhibited significantly superior performance compared to benchmarked algorithms (p ≤ 0.01 post Bonferroni correction)., Conclusion: ARD-Net provides a robust framework to generate SC-PET from LC-PET images. ARD-Net generated SC-PET images exhibited superior performance compared other DL and non-DL approaches in terms of image-fidelity based metrics, task-based segmentation metrics, and minimal bias in terms of task-based quantification performance for preclinical PET imaging., (© 2024 The Authors. Medical Physics published by Wiley Periodicals LLC on behalf of American Association of Physicists in Medicine.)

Segmentation of organs and lesions could be employed for the express purpose of dosimetry in nuclear medicine, assisted image interpretations, and mass image processing studies. Deep leaning created liver and liver lesion segmentation on clinical 3D MRI data has not been fully addressed in previous experiments. To this end, the required data were collected from 128 patients, including their T1w and T2w MRI images, and ground truth labels of the liver and liver lesions were generated. The collection of 110 T1w-T2w MRI image sets was divided, with 94 designated for training and 16 for validation. Furthermore, 18 more datasets were separately allocated for use as hold-out test datasets. The T1w and T2w MRI images were preprocessed into a two-channel format so that they were used as inputs to the deep learning model based on the Isensee 2017 network. To calculate the final Dice coefficient of the network performance on test datasets, the binary average of T1w and T2w predicted images was used. The deep learning model could segment all 18 test cases, with an average Dice coefficient of 88% for the liver and 53% for the liver tumor. Liver segmentation was carried out with rather a high accuracy; this could be achieved for liver dosimetry during systemic or selective radiation therapies as well as for attenuation correction in PET/MRI scanners. Nevertheless, the delineation of liver lesions was not optimal; therefore, tumor detection was not practical by the proposed method on clinical data., (© 2024. Australasian College of Physical Scientists and Engineers in Medicine.)

Crowd flow prediction has been studied for a variety of purposes, ranging from the private sector such as location selection of stores according to the characteristics of commercial districts and customer-tailored marketing to the public sector for social infrastructure design such as transportation networks. Its importance is even greater in light of the spread of contagious diseases such as COVID-19. In many cases, crowd flow can be divided into subgroups by common characteristics such as gender, age, location type, etc. If we use such hierarchical structure of the data effectively, we can improve prediction accuracy of crowd flow for subgroups. But the existing prediction models do not consider such hierarchical structure of the data. In this study, we propose a deep learning model based on global-local structure of the crowd flow data, which utilizes the overall(global) and subdivided by the types of sites(local) crowd flow data simultaneously to predict the crowd flow of each subgroup. The experiment result shows that the proposed model improves the prediction accuracy of each sub-divided subgroup by 5.2% (Table 5 Cat #9)-45.95% (Table 11 Cat #5), depending on the data set. This result comes from the comparison with the related works under the same condition that use target category data to predict each subgroup. In addition, when we refine the global data composition by considering the correlation between subgroups and excluding low correlated subgroups, the prediction accuracy is further improved by 5.6-48.65%., (© 2024. The Author(s).)

In this study, we developed a method for generating quasi-material decomposition (quasi-MD) images from single-energy computed tomography (SECT) images using a deep convolutional neural network (DCNN). Our aim was to improve the detection of cholesterol gallstones and to determine the clinical utility of quasi-MD images. Four thousand pairs of virtual monochromatic images (70 keV) and MD images (fat/water) of the same section, obtained via dual-energy computed tomography (DECT), were used to train the DCNN. The trained DCNN can automatically generate quasi-MD images from the SECT images. Additional SECT images were obtained from 70 patients (40 with and 30 without cholesterol gallstones) to generate quasi-MD images for testing. The presence of gallstones in this dataset was confirmed by ultrasonography. We conducted a receiver operating characteristic (ROC) observer study with three radiologists to validate the clinical utility of the quasi-MD images for detecting cholesterol gallstones. The mean area under the ROC curve for the detection of cholesterol gallstones improved from 0.867 to 0.921 (p = 0.001) when quasi-MD images were added to SECT images. The clinical utility of quasi-MD imaging for detecting cholesterol gallstones was showed. This study demonstrated that the lesion detection capability of images obtained from SECT can be improved using a DCNN trained with DECT images obtained using high-end computed tomography systems., (© 2024. The Author(s), under exclusive licence to Japanese Society of Radiological Technology and Japan Society of Medical Physics.)

Purpose: The efficacy of adjuvant chemotherapy in elderly breast cancer patients is currently controversial. This study aims to provide personalized adjuvant chemotherapy recommendations using deep learning (DL)., Methods: Six models with various causal inference approaches were trained to make individualized chemotherapy recommendations. Patients who received actual treatment recommended by DL models were compared with those who did not. Inverse probability treatment weighting (IPTW) was used to reduce bias. Linear regression, IPTW-adjusted risk difference (RD), and SurvSHAP(t) were used to interpret the best model., Results: A total of 5352 elderly breast cancer patients were included. The median (interquartile range) follow-up time was 52 (30-80) months. Among all models, the balanced individual treatment effect for survival data (BITES) performed best. Treatment according to following BITES recommendations was associated with survival benefit, with a multivariate hazard ratio (HR) of 0.78 (95% confidence interval (CI): 0.64-0.94), IPTW-adjusted HR of 0.74 (95% CI: 0.59-0.93), RD of 12.40% (95% CI: 8.01-16.90%), IPTW-adjusted RD of 11.50% (95% CI: 7.16-15.80%), difference in restricted mean survival time (dRMST) of 12.44 (95% CI: 8.28-16.60) months, IPTW-adjusted dRMST of 7.81 (95% CI: 2.93-11.93) months, and p value of the IPTW-adjusted Log-rank test of 0.033. By interpreting BITES, the debiased impact of patient characteristics on adjuvant chemotherapy was quantified, which mainly included breast cancer subtype, tumor size, number of positive lymph nodes, TNM stages, histological grades, and surgical type., Conclusion: Our results emphasize the potential of DL models in guiding adjuvant chemotherapy decisions for elderly breast cancer patients., (© 2024. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.)

Objectives: This study aimed to use a deep learning (DL) approach for the automatic identification of the ridge deficiency around dental implants based on an image slice from cone-beam computerized tomography (CBCT)., Materials and Methods: Single slices crossing the central long-axis of 630 mandibular and 845 maxillary virtually placed implants (4-5 mm diameter, 10 mm length) in 412 patients were used. The ridges were classified based on the intraoral bone-implant support and sinus floor location. The slices were either preprocessed by alveolar ridge homogenizing prior to DL (preprocessed) or left unpreprocessed. A convolutional neural network with ResNet-50 architecture was employed for DL., Results: The model achieved an accuracy of >98.5% on the unpreprocessed image slices and was found to be superior to the accuracy observed on the preprocessed slices. On the mandible, model accuracy was 98.91 ± 1.45%, and F1 score, a measure of a model's accuracy in binary classification tasks, was lowest (97.30%) on the ridge with a combined horizontal-vertical defect. On the maxilla, model accuracy was 98.82 ± 1.11%, and the ridge presenting an implant collar-sinus floor distance of 5-10 mm with a dehiscence defect had the lowest F1 score (95.86%). To achieve >90% model accuracy, ≥441 mandibular slices or ≥592 maxillary slices were required., Conclusions: The ridge deficiency around dental implants can be identified using DL from CBCT image slices without the need for preprocessed homogenization. The model will be further strengthened by implementing more clinical expertise in dental implant treatment planning and incorporating multiple slices to classify 3-dimensional implant-ridge relationships., (© 2023 Wiley Periodicals LLC.)

In the recent years, the academic paper influence analysis has been widely studied due to its potential applications in the multiple areas of science information metric and retrieval. By identifying the academic influence of papers, authors, etc., we can directly support researchers to easily reach academic papers. These recommended candidate papers are not only highly relevant with their desired research topics but also highly-attended by the research community within these topics. For very recent years, the rapid developments of academic networks, like Google Scholar, Research Gate, CiteSeerX, etc., have significantly boosted the number of new published papers annually. It also helps to strengthen the borderless cooperation between researchers who are interested on the same research topics. However, these current academic networks still lack the capabilities of provisioning researchers deeper into most-influenced papers. They also largely ignore quite/irrelevant papers, which are not fully related with their current interest topics. Moreover, the distributions of topics within these academic papers are considered as varying and it is difficult to extract the main concentrated topics in these papers. Thus, it leads to challenges for researchers to find their appropriated/high-qualified reference resources while doing researches. To overcome this limitation, in this paper, we proposed a novel approach of paper influence analysis through their content-based and citation relationship-based analyses within the biographical network. In order to effectively extract the topic-based relevance from papers, we apply the integrated graph-based citation relationship analysis with topic modeling approach to automatically learn the distributions of keyword-based labeled topics in forms of unsupervised learning approach, named as TopCite. Then, we base on the constructed graph-based paper–topic structure to identify their relevancy levels. Upon the identified relevancy levels between papers, we can support for improving the accuracy performance of other bibliographic network mining tasks, such as paper similarity measurement, recommendation, etc. Extensive experiments in real-world AMiner bibliographic dataset demonstrate the effectiveness of our proposed ideas in this paper. [ABSTRACT FROM AUTHOR]

Benefiting from the growth of the bandwidth, Terahertz (THz) communication can support the new application with explosive requirements of the ultra-high-speed rates for future 6G wireless systems. In order to compensate for the path loss of high frequency, massive Multiple-Input Multiple-Output (MIMO) can be utilized for high array gains by beamforming. However, the existing THz communication with massive MIMO has remarkably high energy consumption because a large number of analog phase shifters should be used to realize the analog beamforming. To solve this problem, a Reconfigurable Intelligent Surface (RIS) based hybrid precoding architecture for THz communication is developed in this paper, where the energy-hungry phased array is replaced by the energy-efficient RIS to realize the analog beamforming of the hybrid precoding. Then, based on the proposed RIS-based architecture, a sum-rate maximization problem for hybrid precoding is investigated. Since the phase shifts implemented by RIS in practice are often discrete, this sum-rate maximization problem with a non-convex constraint is challenging. Next, the sum-rate maximization problem is reformulated as a parallel Deep Neural Network (DNN) based classification problem, which can be solved by the proposed low-complexity Deep Learning based Multiple Discrete Classification (DL-MDC) hybrid precoding scheme. Finally, we provide numerous simulation results to show that the proposed DL-MDC scheme works well both in the theoretical Saleh-Valenzuela channel model and practical 3GPP channel model. Compared with existing iterative search algorithms, the proposed DL-MDC scheme significantly reduces the runtime with a negligible performance loss.

Music generation using deep learning has received considerable attention in recent years. Researchers have developed various generative models capable of imitating musical conventions, comprehending the musical corpora, and generating new samples based on the learning outcome. Although the samples generated by these models are persuasive, they often lack musical structure and creativity. For instance, a vanilla end-to-end approach, which deals with all levels of music representation at once, does not offer human-level control and interaction during the learning process, leading to constrained results. Indeed, music creation is a recurrent process that follows some principles by a musician, where various musical features are reused or adapted. On the other hand, a musical piece adheres to a musical style, breaking down into precise concepts of timbre style, performance style, composition style, and the coherency between these aspects. Here, we study and analyze the current advances in music generation using deep learning models through different criteria. We discuss the shortcomings and limitations of these models regarding interactivity and adaptability. Finally, we draw the potential future research direction addressing multi-agent systems and reinforcement learning algorithms to alleviate these shortcomings and limitations.

The research groups in computer vision, graphics, and machine learning have dedicated a substantial amount of attention to the areas of 3D object reconstruction, augmentation, and registration. Deep learning is the predominant method used in artificial intelligence for addressing computer vision challenges. However, deep learning on three-dimensional data presents distinct obstacles and is now in its nascent phase. There have been significant advancements in deep learning specifically for three-dimensional data, offering a range of ways to address these issues. This study offers a comprehensive examination of the latest advancements in deep learning methodologies. We examine many benchmark models for the tasks of 3D object registration, augmentation, and reconstruction. We thoroughly analyse their architectures, advantages, and constraints. In summary, this report provides a comprehensive overview of recent advancements in three-dimensional deep learning and highlights unresolved research areas that will need to be addressed in the future. [ABSTRACT FROM AUTHOR]

In recent years, a variety of deep learning networks for cardiac MRI (CMR) segmentation have been developed and analyzed. However, nearly all of them are focused on cine CMR under breathold. In this work, accuracy of deep learning methods is assessed for volumetric analysis (via segmentation) of the left ventricle in real-time free-breathing CMR at rest and under exercise stress. Data from healthy volunteers (n = 15) for cine and real-time free-breathing CMR at rest and under exercise stress were analyzed retrospectively. Exercise stress was performed using an ergometer in the supine position. Segmentations of two deep learning methods, a commercially available technique (comDL) and an openly available network (nnU-Net), were compared to a reference model created via the manual correction of segmentations obtained with comDL. Segmentations of left ventricular endocardium (LV), left ventricular myocardium (MYO), and right ventricle (RV) are compared for both end-systolic and end-diastolic phases and analyzed with Dice's coefficient. The volumetric analysis includes the cardiac function parameters LV end-diastolic volume (EDV), LV end-systolic volume (ESV), and LV ejection fraction (EF), evaluated with respect to both absolute and relative differences. For cine CMR, nnU-Net and comDL achieve Dice's coefficients above 0.95 for LV and 0.9 for MYO, and RV. For real-time CMR, the accuracy of nnU-Net exceeds that of comDL overall. For real-time CMR at rest, nnU-Net achieves Dice's coefficients of 0.94 for LV, 0.89 for MYO, and 0.90 for RV and the mean absolute differences between nnU-Net and the reference are 2.9 mL for EDV, 3.5 mL for ESV, and 2.6% for EF. For real-time CMR under exercise stress, nnU-Net achieves Dice's coefficients of 0.92 for LV, 0.85 for MYO, and 0.83 for RV and the mean absolute differences between nnU-Net and reference are 11.4 mL for EDV, 2.9 mL for ESV, and 3.6% for EF. Deep learning methods designed or trained for cine CMR segmentation can perform well on real-time CMR. For real-time free-breathing CMR at rest, the performance of deep learning methods is comparable to inter-observer variability in cine CMR and is usable for fully automatic segmentation. For real-time CMR under exercise stress, the performance of nnU-Net could promise a higher degree of automation in the future., (© 2024. The Author(s).)

Purpose: The quality of ultrasound images is degraded by speckle and Gaussian noises. This study aims to develop a deep-learning (DL)-based filter for ultrasound image denoising., Methods: A novel DL-based filter using adaptive residual (AdaRes) learning was proposed. Five image quality metrics (IQMs) and 27 radiomics features were used to evaluate denoising results. The effect of our proposed filter, AdaRes, on four pre-trained convolutional neural network (CNN) classification models and three radiologists was assessed., Results: AdaRes filter was tested on both natural and ultrasound image databases. IQMs results indicate that AdaRes could remove noises in three different noise levels with the highest performances. In addition, a radiomics study proved that AdaRes did not distort tissue textures and it could preserve most radiomics features. AdaRes could also improve the performance classification using CNNs in different settings. Finally, AdaRes also improved the mean overall performance (AUC) of three radiologists from 0.494 to 0.702 in the classification of benign and malignant lesions., Conclusions: AdaRes filtered out noises on ultrasound images more effectively and can be used as an auxiliary preprocessing step in computer-aided diagnosis systems. Radiologists may use it to remove unwanted noises and improve the ultrasound image quality before the interpretation., (© 2023 Wiley Periodicals LLC.)

The breathtaking economic development put a heavy toll on ecology, especially on water pollution. Efficient water resource management has a long-term influence on the sustainable development of the economy and society. Economic development and ecology preservation are tangled together, and the growth of one is not possible without the other. Deep learning (DL) is ubiquitous in autonomous driving, medical imaging, speech recognition, etc. The spectacular success of deep learning comes from its power of richer representation of data. In view of the bright prospects of DL, this review comprehensively focuses on the development of DL applications in water resources management, water environment protection, and water ecology. First, the concept and modeling steps of DL are briefly introduced, including data preparation, algorithm selection, and model evaluation. Finally, the advantages and disadvantages of commonly used algorithms are analyzed according to their structures and mechanisms, and recommendations on the selection of DL algorithms for different studies, as well as prospects for the application and development of DL in water science are proposed. This review provides references for solving a wider range of water-related problems and brings further insights into the intelligent development of water science., (© 2024. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)

Background: Attenuation and scatter correction is crucial for quantitative positron emission tomography (PET) imaging. Direct attenuation correction (AC) in the image domain using deep learning approaches has been recently proposed for combined PET/MR and standalone PET modalities lacking transmission scanning devices or anatomical imaging., Purpose: In this study, different input settings were considered in the model training to investigate deep learning-based AC in the image space., Methods: Three different deep learning methods were developed for direct AC in the image space: (i) use of non-attenuation-corrected PET images as input (NonAC-PET), (ii) use of attenuation-corrected PET images with a simple two-class AC map (composed of soft-tissue and background air) obtained from NonAC-PET images (PET segmentation-based AC [SegAC-PET]), and (iii) use of both NonAC-PET and SegAC-PET images in a Double-Channel fashion to predict ground truth attenuation corrected PET images with Computed Tomography images (CTAC-PET). Since a simple two-class AC map (generated from NonAC-PET images) can easily be generated, this work assessed the added value of incorporating SegAC-PET images into direct AC in the image space. A 4-fold cross-validation scheme was adopted to train and evaluate the different models based using 80 brain 18 F-Fluorodeoxyglucose PET/CT images. The voxel-wise and region-wise accuracy of the models were examined via measuring the standardized uptake value (SUV) quantification bias in different regions of the brain., Results: The overall root mean square error (RMSE) for the Double-Channel setting was 0.157 ± 0.08 SUV in the whole brain region, while RMSEs of 0.214 ± 0.07 and 0.189 ± 0.14 SUV were observed in NonAC-PET and SegAC-PET models, respectively. A mean SUV bias of 0.01 ± 0.26% was achieved by the Double-Channel model regarding the activity concentration in cerebellum region, as opposed to 0.08 ± 0.28% and 0.05 ± 0.28% SUV biases for the network that uniquely used NonAC-PET or SegAC-PET as input, respectively. SegAC-PET images with an SUV bias of -1.15 ± 0.54%, served as a benchmark for clinically accepted errors. In general, the Double-Channel network, relying on both SegAC-PET and NonAC-PET images, outperformed the other AC models., Conclusion: Since the generation of two-class AC maps from non-AC PET images is straightforward, the current study investigated the potential added value of incorporating SegAC-PET images into a deep learning-based direct AC approach. Altogether, compared with models that use only NonAC-PET and SegAC-PET images, the Double-Channel deep learning network exhibited superior attenuation correction accuracy., (© 2024 American Association of Physicists in Medicine.)

Digital reconstruction of neuronal structures from 3D neuron microscopy images is critical for the quantitative investigation of brain circuits and functions. Currently, neuron reconstructions are mainly obtained by manual or semiautomatic methods. However, these ways are labor-intensive, especially when handling the huge volume of whole brain microscopy imaging data. Here, we present a deep-learning-based neuron morphology analysis toolbox (DNeuroMAT) for automated analysis of neuron microscopy images, which consists of three modules: neuron segmentation, neuron reconstruction, and neuron critical points detection., (© 2024. The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature.)

Objective: To enable the intelligent diagnosis of a variety of common Electrocardiogram (ECG), we investigate the deep learning-based ECG diagnosis system., Methods: From January 2015 to December 2019, four consecutive years of 100,120 conventional 12-lead ECG data were collected in our hospital. Utilizing this dataset, we constructed a deep learning model designed to intelligently diagnose prevalent ECG anomalies by employing a multi-task learning framework. The system performance was evaluated using various metrics, including sensitivity, specificity, negative predictive value, positive predictive value, and so forth. Additionally, we employed an ECG intelligent diagnostic platform for clinical application to undertake real-time online analysis of 2500 conventional 12-lead ECG samples in June 2020, aiming to validate our model. At this stage, we compared the performance of our model against the traditional manual identification method., Results: The efficacy of the ECG intelligent diagnostic model was notably high for common and straightforward ECG patterns, such as sinus rhythm (F1 = 98.01%), sinus tachycardia (F1 = 96.26%), sinus bradycardia (F1 = 94.88%), and a normal electrocardiogram (F1 = 91.71%), as well as for Premature Ventricular Contractions (F1 = 91.62%). Nevertheless, when diagnosing rarer and more intricate ECG anomalies, the system requires an increased number of samples to refine the deep learning models. During the validation stage, our model exhibited better efficiency in terms of accuracy, labor time and labor cost when compared to the manual identification approach., Conclusions: Our deep learning-driven intelligent ECG diagnostic model clearly demonstrates significant clinical utility. The integrated artificial intelligence diagnosis system not only has the potential to augment physicians in their diagnostic processes but also offers a viable avenue to reduce associated labor costs., (© 2023 Wiley Periodicals LLC.)

Background: Low-dose positron emission tomography (LD-PET) imaging is commonly employed in preclinical research to minimize radiation exposure to animal subjects. However, LD-PET images often exhibit poor quality and high noise levels due to the low signal-to-noise ratio. Deep learning (DL) techniques such as generative adversarial networks (GANs) and convolutional neural network (CNN) have the capability to enhance the quality of images derived from noisy or low-quality PET data, which encodes critical information about radioactivity distribution in the body., Purpose: Our objective was to optimize the image quality and reduce noise in preclinical PET images by utilizing the sinogram domain as input for DL models, resulting in improved image quality as compared to LD-PET images., Methods: A GAN and CNN model were utilized to predict high-dose (HD) preclinical PET sinograms from the corresponding LD preclinical PET sinograms. In order to generate the datasets, experiments were conducted on micro-phantoms, animal subjects (rats), and virtual simulations. The quality of DL-generated images was weighted by performing the following quantitative measures: structural similarity index measure (SSIM), root mean squared error (RMSE), peak signal-to-noise ratio (PSNR), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR). Additionally, DL input and output were both subjected to a spatial resolution calculation of full width half maximum (FWHM) and full width tenth maximum (FWTM). DL outcomes were then compared with the conventional denoising algorithms such as non-local means (NLM), block-matching, and 3D filtering (BM3D)., Results: The DL models effectively learned image features and produced high-quality images, as reflected in the quantitative metrics. Notably, the FWHM and FWTM values of DL PET images exhibited significantly improved accuracy compared to LD, NLM, and BM3D PET images, and just as precise as HD PET images. The MSE loss underscored the excellent performance of the models, indicating that the models performed well. To further improve the training, the generator loss (G loss) was increased to a value higher than the discriminator loss (D loss), thereby achieving convergence in the GAN model., Conclusions: The sinograms generated by the GAN network closely resembled real HD preclinical PET sinograms and were more realistic than LD. There was a noticeable improvement in image quality and noise factor in the predicted HD images. Importantly, DL networks did not fully compromise the spatial resolution of the images., (© 2023 American Association of Physicists in Medicine.)

The detection of concealed suspicious objects in public places is a critical issue and a popular research topic. Terahertz (THz) imaging technology, as an emerging detection method, can penetrate materials without emitting ionizing radiation, providing a new approach to detecting concealed suspicious objects. This study focuses on the detection of concealed suspicious objects wrapped in different materials such as polyethylene and kraft paper, including items like scissors, pistols, and blades, using THz imaging technology. To address issues such as the lack of texture details in THz images and the contour similarity of different objects, which can lead to missed detections and false alarms, we propose a THz concealed suspicious object detection model based on SMR–YOLO (SPD_Mobile + RFB + YOLO). This model, based on the MobileNext network, introduces the spatial-to-depth convolution (SPD-Conv) module to replace the backbone network, reducing computational and parameter load. The inclusion of the receptive field block (RFB) module, which uses a multi-branch structure of dilated convolutions, enhances the network's depth features. Using the EIOU loss function to assess the accuracy of predicted box localization further optimizes convergence speed and localization accuracy. Experimental results show that the improved model achieved mAP@0.5 and mAP@0.5:0.95 scores of 98.9% and 89.4%, respectively, representing improvements of 0.2% and 1.8% over the baseline model. Additionally, the detection speed reached 108.7 FPS, an improvement of 23.2 FPS over the baseline model. The model effectively identifies concealed suspicious objects within packages, offering a novel approach for detection in public places. [ABSTRACT FROM AUTHOR]

The paper classification method aims to correctly divide the paper data according to the similarity of its content. However, how to accurately classify according to the content expressed in the paper has always been a problem that various classification algorithms need to face. At present, there is a kind of paper classification method based on deep learning and implemented by the encoder‐decoder structure. This method inputs the words from a large number of papers into encoder, after calculating by NN (neural network) algorithm, the similarity degree of different papers is compared to achieve the purpose of classification. However, this type of method only considers the similarity between words, a NN algorithm can only calculate a large number of word information once, and it cannot find the regularity of classification through word information. But it has a difference with the similarity of the content. This paper starts from the perspective of considering the content, its label information is extracted, and the input vector of encoder‐decoder structure is formed with labels and words. This improves the original paper classification method based on encoder‐decoder structure. Firstly, the label information is based on the content, which can reflect the content of the paper. Secondly, the classification method which combines label information and word information can reflect the content of the paper comprehensively. Thirdly, the label information is independent of word information and NN algorithm is used separately to make this part of the content more consistent in the encoder‐decoder structure. Finally, the label information and the word information are combined, respectively, with the output values obtained by different NN algorithms to realize the classification of the content. This paper proves the effectiveness of the proposed method by evaluating the paper data in web of science and obtaining relevant experimental results. [ABSTRACT FROM AUTHOR]

The 2023 edition of the AIxIA Conference, held in Rome, brought together a large number of researchers and practitioners to discuss the most recent and important advancements in Artificial Intelligence (AI). The conference featured 19 workshops, organized by 77 experts, attracting 248 submissions and resulting in 16 proceedings. This special issue presents extended versions of selected papers initially showcased at these workshops. Each paper underwent rigorous review and represents a diverse array of topics, reflecting the multifaceted nature of the Italian AI community. The topics covered include ethical foundations to symbiotic AI, symbolic knowledge extraction from black-box models, creative influence prediction using graph theory, AI approaches to multidimensional poverty prediction, an assessment of AI-based supports for informal caregivers, deep learning-based EEG denoising, AI-assisted board-game-based learning, large language models for assessment and feedback in higher education, geometric reasoning in the Traveling Salesperson Problem, defeasible reasoning in weighted knowledge bases, and conditional computation in neural networks. These contributions demonstrate the innovative and interdisciplinary research within the AI community, offering valuable insights and advancing the field. [ABSTRACT FROM AUTHOR]

To explore whether the brain contains pattern differences in the rock–paper–scissors (RPS) imagery task, this paper attempts to classify this task using fNIRS and deep learning. In this study, we designed an RPS task with a total duration of 25 min and 40 s, and recruited 22 volunteers for the experiment. We used the fNIRS acquisition device (FOIRE-3000) to record the cerebral neural activities of these participants in the RPS task. The time series classification (TSC) algorithm was introduced into the time-domain fNIRS signal classification. Experiments show that CNN-based TSC methods can achieve 97% accuracy in RPS classification. CNN-based TSC method is suitable for the classification of fNIRS signals in RPS motor imagery tasks, and may find new application directions for the development of brain–computer interfaces (BCI).

To improve the accuracy of user implicit rating prediction, we combine the traditional latent factor model (LFM) and bidirectional gated recurrent unit neural network (BiGRU) model to propose a hybrid model that deeply mines the latent semantics in the unstructured content of the text and generates a more accurate rating matrix. First, we utilize the user’s historical behavior (favorites records) to build a user rating matrix and decompose the matrix to obtain the latent factor vectors of users and literature. We also apply the BERT model for word embedding of the research papers to obtain the sequence of word vectors. Then, we apply the BiGRU with the user attention mechanism to mine the research paper textual content and to generate the new literature latent feature vectors that are used to replace the original literature latent factor vectors decomposed from the rating matrix. Finally, a new rating matrix is generated to obtain users’ ratings of noninteractive research papers and to generate the recommendation list according to the user latent factor vector. We design experiments on the real datasets and verify that the research paper recommendation model is superior to traditional recommendation models in terms of precision, recall, F1-value, coverage, popularity and diversity.

Digital systems storing cadastral data in vector format are considered effective due to their ability of offering interactive services to citizens and other land-related systems. The adoption of such systems is ubiquitous, but when adopted, they create two non-compatible systems with paper-based cadastral systems whose information needs to be digitised. This study proposes a new approach that is fast and accurate for transforming paper-based cadastral data into digital systems. The proposed method involves deep-learning techniques of the LCNN and ResNet-50 for detecting cadastral parcels and their numbers, respectively, from the cadastral plans. It also contains four functions defined to speed up transformations and compilations of the cadastral plan's data in digital systems. The LCNN is trained and validated with 968 samples. The ResNet-50 is trained and validated with 106,000 samples. The Structural-Average-Precision ( sAP 10 ) achieved with the LCNN was 0.9057. The Precision, Recall and F1-Score achieved with the ResNet-50 were 0.9650, 0.9648 and 0.9649, respectively. These results confirmed that the new method is accurate enough for implementation, and we tested it with a huge set of data from Tanzania. Its performance from the experimented data shows that the proposed method could effectively transform paper-based cadastral data into digital systems. [ABSTRACT FROM AUTHOR]

Machine learning (ML), a form of artificial intelligence (AI), is being increasingly employed in neurology. Reported performance metrics often match or exceed the efficiency of average clinicians. The neurologist is easily baffled by the underlying concepts and terminologies associated with ML studies. The superlative performance metrics of ML algorithms often hide the opaque nature of its inner workings. Questions regarding ML model's interpretability and reproducibility of its results in real-world scenarios, need emphasis. Given an abundance of time and information, the expert clinician should be able to deliver comparable predictions to ML models, a useful benchmark while evaluating its performance. Predictive performance metrics of ML models should not be confused with causal inference between its input and output. ML and clinical gestalt should compete in a randomized controlled trial before they can complement each other for screening, triaging, providing second opinions and modifying treatment.

These data were used to build an object detection model to locate Relict Charcoal Hearths (RCH) as described in the paper “When Computers Dream of Charcoal: Using Deep Learning, Open Tools and Open Data to Identify Relict Charcoal Hearths in and around State Game Lands in Pennsylvania” [1]. This is the second grouping of data for the paper above. The first grouping is also available in this journal, see “Geospatial and image data from the “When Computers Dream of Charcoal: Using Deep Learning, Open Tools and Open Data to Identify Relict Charcoal Hearths in and around State Game Lands in Pennsylvania” paper” [2]. These files consist of: JPEGs representing tiles of larger Slope TIFF files derived from LiDAR for the State Game Lands (SGL) of Pennsylvania, United States [3456]. A subset of these tiles was used to train the model. A Shapefile of points of known relict charcoal hearths (RCH). XML files representing the pixel points of known RCHs on JPEG files used for training. Jupyter notebooks of programs used to prepare data and train a Mask R-CNN model. The Mask R-CNN model H5 file. Shapefile and GeoJSON of object detection results from the model showing locations of possible RCH in all SGLs. XML files representing the pixel points of predicted RCH on JPEG files used for predictions. GeoJSON of results using cluster analysis. These data are stored onzenodo.org. The programs are stored on Github.com.

These data were used to build an object detection model to locate Relict Charcoal Hearths (RCH) as described in the paper “When Computers Dream of Charcoal: Using Deep Learning, Open Tools and Open Data to Identify Relict Charcoal Hearths in and around State Game Lands in Pennsylvania” [1]. This is the first grouping of data for the paper above. The second grouping is also available in this journal, see “Object detection model, image data and results from the “When Computers Dream of Charcoal: Using Deep Learning, Open Tools and Open Data to Identify Relict Charcoal Hearths in and around State Game Lands in Pennsylvania” paper”. These files consist of georeferenced Digital Elevation Model, Hillshade and Slope files derived from LiDAR for the State Game Lands (SGL). Included is a Shapefile and GeoJSON of State Game Land borders as well as the program used for downloading the LiDAR files. These data are stored on 'Zenodo.org'.

Medical imaging is currently using artificial intelligence-based technologies to aid evaluate diagnostic information images, particularly in enforcing kidney stones. Artificial intelligence technology continues to develop, many studies show that deep learning is more widely used compared to traditional machine learning, so an Artificial intelligence system is needed to assist the accuracy of health diagnoses, thus helping in the field of radiology health. The aim of the research is to use artificial intelligence with deep learning models to help detect abnormalities in the kidneys. This research method is a literature review of Scopus data related to deep learning in medical imaging in detecting kidney stones. The results of using Artificial Intelligence in medical imaging can be used in diagnosing diseases including detecting Covid-19, musculoskeletal, calcium scores on Cardiac CT, liver tumors, urinary tract lesions, examination of the abdomen and kidney stones. Utilization of Artificial Intelligence in detecting kidney stones can be done with various classification models including XResNet-50, ExDark19, CystoNet, CNN, ANN. Using the right model and having a high accuracy value can help radiologists to accurately detect kidney stones.

Objectives: Segmentation is crucial in medical imaging. Deep learning based on convolutional neural networks showed promising results. However, the absence of large-scale datasets and a high degree of inter- and intra-observer variations pose a bottleneck. Crowdsourcing might be an alternative, as many non-experts provide references. We aim to compare different types of crowdsourcing for medical image segmentation., Methods: We develop a crowdsourcing platform that integrates citizen science (incentive: participating in the research), paid microtask (incentive: financial reward), and gamification (incentive: entertainment). For evaluation, we choose the use case of sclera segmentation in fundus images as a proof-of-concept and analyze the accuracy of crowdsourced masks and the generalization of learning models trained with crowdsourced masks., Results: The developed platform is suited for the different types of crowdsourcing and offers an easy and intuitive way to implement crowdsourcing studies. Regarding the proof-of-concept study, citizen science, paid microtask, and gamification yield a median F-score of 82.2, 69.4, and 69.3 % compared to expert-labeled ground truth, respectively. Generating consensus masks improves the gamification masks (78.3 %). Despite the small training data (50 images), deep learning reaches median F-scores of 80.0, 73.5, and 76.5 % for citizen science, paid microtask, and gamification, respectively, indicating sufficient generalizability., Conclusions: As the platform has proven useful, we aim to make it available as open-source software for other researchers., (© 2023 the author(s), published by De Gruyter, Berlin/Boston.)

Objectives: This study aims at a fully automatic pipeline for measuring the magnetic resonance parkinsonism index (MRPI) using deep learning methods., Methods: MRPI is defined as the product of the pons area to the midbrain area ratio and the middle cerebellar peduncle (MCP) width to the superior cerebellar peduncle (SCP) width ratio. In our proposed pipeline, we first used nnUNet to segment the brainstem and then employed HRNet to identify two key boundary points so as to sub-divide the whole brainstem into midbrain and pons. HRNet was also employed to predict the MCP endpoints for measuring the MCP width. Finally, we segmented the SCP on an oblique coronal plane and calculated its width. A total of 400 T1-weighted magnetic resonance images (MRIs) were used to train the nnUNet and HRNet models. Five-fold cross-validation was conducted to evaluate our proposed pipeline's performance on the training dataset. We also evaluated the performance of our proposed pipeline on three external datasets. Two of them had two raters manually measuring the MRPI values, providing insights into automatic accuracy versus inter-rater variability., Results: We obtained average absolute percentage errors (APEs) of 17.21%, 18.17%, 20.83%, and 22.83% on the training dataset and the three external validation datasets, while the inter-rater average APE measured on the first two external validation datasets was 11.31%. Our proposed pipeline significantly improved the MRPI quantification accuracy over a representative state-of-the-art traditional approach (p < 0.001)., Conclusion: The proposed automatic pipeline can accurately predict MRPI that is comparable with manual measurement., Clinical Relevance Statement: This study presents an automated magnetic resonance parkinsonism index measurement tool that can analyze large amounts of magnetic resonance images, enhance the efficiency of Parkinsonism-Plus syndrome diagnosis, reduce the workload of clinicians, and minimize the impact of human factors on diagnosis., Key Points: • We propose an automatic pipeline for measuring the magnetic resonance parkinsonism index from magnetic resonance images. • The effectiveness of the proposed pipeline is successfully established on multiple datasets and comparisons with inter-rater measurements. • The proposed pipeline significantly outperforms a state-of-the-art quantification approach, being much closer to ground truth., (© 2023. The Author(s), under exclusive licence to European Society of Radiology.)

Background: Pathological maxillary sinus would affect implant treatment and even result in failure of maxillary sinus lift and implant surgery. However, the maxillary sinus abnormalities are challenging to be diagnosed through CBCT images, especially for young dentists or dentists in grassroots medical institutions without systematical education of general medicine., Objectives: To develop a deep-learning-based screening model incorporating object detection and 'straight-forward' classification strategy to screen out maxillary sinus abnormalities on CBCT images., Methods: The large area of background noise outside maxillary sinus would affect the generalisation and prediction accuracy of the model, and the diversity and imbalanced distribution of imaging manifestations may bring challenges to intellectualization. Thus we adopted an object detection to limit model's observation zone and 'straight-forward' classification strategy with various tuning methods to adapt to dental clinical need and extract typical features of diverse manifestations so that turn the task into a 'normal-or-not' classification., Results: We successfully constructed a deep-learning model consist of well-trained detector and diagnostor module. This model achieved ideal AUROC and AUPRC of 0.953 and 0.887, reaching more than 90% accuracy at optimal cut-off. McNemar and Kappa test verified no statistical difference and high consistency between the prediction and ground truth. Dentist-model comparison test showed the model's statistically higher diagnostic performance than dental students. Visualisation method confirmed the model's effectiveness in region recognition and feature extraction., Conclusion: The deep-learning model incorporating object detection and straightforward classification strategy could achieve satisfying predictive performance for screening maxillary sinus abnormalities on CBCT images., (© 2023 John Wiley & Sons Ltd.)

Background: The application of cone-beam computed tomography (CBCT) in image-guided radiotherapy and adaptive radiotherapy remains limited due to its poor image quality., Purpose: In this study, we aim to develop a deep learning framework to generate high-quality CBCT images for therapeutic applications., Methods: The synthetic CT (sCT) generation from the CBCT was proposed using a transformer-based network with a hybrid loss function. The network was trained and validated using the data from 176 patients to produce a general model that can be extensively applied to enhance CBCT images. After the first therapy, each patient can receive paired CBCT/planning CT (pCT) scans, and the obtained data were used to fine-tune the general model for further improvement. For subsequent treatment, a patient-specific, personalized model was made available. In total, 34 patients were examined for general model testing, and another six patients who underwent rescanned pCT scan were used for personalized model training and testing., Results: The general model decreased the mean absolute error (MAE) from 135 HU to 59 HU as compared to the CBCT. The hybrid loss function demonstrated superior performance in CT number correction and noise/artifacts reduction. The proposed transformer-based network also showed superior power in CT number correction compared to the classical convolutional neural network. The personalized model showed improvement based on the general model in some details, and the MAE was reduced from 59 HU (for the general model) to 57 HU (p < 0.05 Wilcoxon signed-rank test)., Conclusion: We established a deep learning framework based on transformer for clinical needs. The deep learning model demonstrated potential for continuous improvement with the help of a suggested personalized training strategy compatible with the clinical workflow., (© 2023 American Association of Physicists in Medicine.)

Purpose: To determine the reliability of an artificial intelligence, deep learning (AI/DL)-based method of chest computer tomography (CT) scan analysis to distinguish pulmonary sarcoidosis from negative lung cancer screening chest CT scans (Lung Imaging Reporting and Data System score 1, Lung-RADS score 1)., Methods: Chest CT scans of pulmonary sarcoidosis were evaluated by a clinician experienced with sarcoidosis and a chest radiologist for clinical and radiologic evidence of sarcoidosis and exclusion of alternative or concomitant pulmonary diseases. The AI/DL based method used an ensemble network architecture combining Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs). The method was applied to 126 pulmonary sarcoidosis and 96 Lung-RADS score 1 CT scans. The analytic approach of training and validation of the AI/DL method used a fivefold cross-validation technique, where 4/5th of the available data set was used to train a diagnostic model and tested on the remaining 1/5th of the data set, and repeated 4 more times with non-overlapping validation/test data. The probability values were used to generate Receiver Operating Characteristic (ROC) curves to assess the model's discriminatory power., Results: The sensitivity, specificity, positive and negative predictive value of the AI/DL method for the 5 folds of the training/validation sets and the entire set of CT scans were all over 94% to distinguish pulmonary sarcoidosis from LUNG-RADS score 1 chest CT scans. The area under the curve for the corresponding ROC curves were all over 97%., Conclusion: This AL/DL model shows promise to distinguish sarcoidosis from alternative pulmonary conditions using minimal radiologic data., (© 2023. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.)

Problem: Implementation of competency-based medical education has necessitated more frequent trainee assessments. Use of simulation as an assessment tool is limited by access to trained examiners, cost, and concerns with interrater reliability. Developing an automated tool for pass/fail assessment of trainees in simulation could improve accessibility and quality assurance of assessments. This study aimed to develop an automated assessment model using deep learning techniques to assess performance of anesthesiology trainees in a simulated critical event., Approach: The authors retrospectively analyzed anaphylaxis simulation videos to train and validate a deep learning model. They used an anaphylactic shock simulation video database from an established simulation curriculum, integrating a convenience sample of 52 usable videos. The core part of the model, developed between July 2019 and July 2020, is a bidirectional transformer encoder., Outcomes: The main outcome was the F1 score, accuracy, recall, and precision of the automated assessment model in analyzing pass/fail of trainees in simulation videos. Five models were developed and evaluated. The strongest model was model 1 with an accuracy of 71% and an F1 score of 0.68., Next Steps: The authors demonstrated the feasibility of developing a deep learning model from a simulation database that can be used for automated assessment of medical trainees in a simulated anaphylaxis scenario. The important next steps are to (1) integrate a larger simulation dataset to improve the accuracy of the model; (2) assess the accuracy of the model on alternative anaphylaxis simulations, additional medical disciplines, and alternative medical education evaluation modalities; and (3) gather feedback from education leadership and clinician educators surrounding the perceived strengths and weaknesses of deep learning models for simulation assessment. Overall, this novel approach for performance prediction has broad implications in medical education and assessment., (Copyright © 2023 by the Association of American Medical Colleges.)

Purpose: To develop a method for building MRI reconstruction neural networks robust to changes in signal-to-noise ratio (SNR) and trainable with a limited number of fully sampled scans., Methods: We propose Noise2Recon, a consistency training method for SNR-robust accelerated MRI reconstruction that can use both fully sampled (labeled) and undersampled (unlabeled) scans. Noise2Recon uses unlabeled data by enforcing consistency between model reconstructions of undersampled scans and their noise-augmented counterparts. Noise2Recon was compared to compressed sensing and both supervised and self-supervised deep learning baselines. Experiments were conducted using retrospectively accelerated data from the mridata three-dimensional fast-spin-echo knee and two-dimensional fastMRI brain datasets. All methods were evaluated in label-limited settings and among out-of-distribution (OOD) shifts, including changes in SNR, acceleration factors, and datasets. An extensive ablation study was conducted to characterize the sensitivity of Noise2Recon to hyperparameter choices., Results: In label-limited settings, Noise2Recon achieved better structural similarity, peak signal-to-noise ratio, and normalized-RMS error than all baselines and matched performance of supervised models, which were trained with 14 × $$ 14\times $$ more fully sampled scans. Noise2Recon outperformed all baselines, including state-of-the-art fine-tuning and augmentation techniques, among low-SNR scans and when generalizing to OOD acceleration factors. Augmentation extent and loss weighting hyperparameters had negligible impact on Noise2Recon compared to supervised methods, which may indicate increased training stability., Conclusion: Noise2Recon is a label-efficient reconstruction method that is robust to distribution shifts, such as changes in SNR, acceleration factors, and others, with limited or no fully sampled training data., (© 2023 International Society for Magnetic Resonance in Medicine.)

Objectives: This study proposes and evaluates a deep learning method to detect pancreatic neoplasms and to identify main pancreatic duct (MPD) dilatation on portal venous computed tomography scans., Materials and Methods: A total of 2890 portal venous computed tomography scans from 9 institutions were acquired, among which 2185 had a pancreatic neoplasm and 705 were healthy controls. Each scan was reviewed by one in a group of 9 radiologists. Physicians contoured the pancreas, pancreatic lesions if present, and the MPD if visible. They also assessed tumor type and MPD dilatation. Data were split into a training and independent testing set of 2134 and 756 cases, respectively.A method to detect pancreatic lesions and MPD dilatation was built in 3 steps. First, a segmentation network was trained in a 5-fold cross-validation manner. Second, outputs of this network were postprocessed to extract imaging features: a normalized lesion risk, the predicted lesion diameter, and the MPD diameter in the head, body, and tail of the pancreas. Third, 2 logistic regression models were calibrated to predict lesion presence and MPD dilatation, respectively. Performance was assessed on the independent test cohort using receiver operating characteristic analysis. The method was also evaluated on subgroups defined based on lesion types and characteristics., Results: The area under the curve of the model detecting lesion presence in a patient was 0.98 (95% confidence interval [CI], 0.97-0.99). A sensitivity of 0.94 (469 of 493; 95% CI, 0.92-0.97) was reported. Similar values were obtained in patients with small (less than 2 cm) and isodense lesions with a sensitivity of 0.94 (115 of 123; 95% CI, 0.87-0.98) and 0.95 (53 of 56, 95% CI, 0.87-1.0), respectively. The model sensitivity was also comparable across lesion types with values of 0.94 (95% CI, 0.91-0.97), 1.0 (95% CI, 0.98-1.0), 0.96 (95% CI, 0.97-1.0) for pancreatic ductal adenocarcinoma, neuroendocrine tumor, and intraductal papillary neoplasm, respectively. Regarding MPD dilatation detection, the model had an area under the curve of 0.97 (95% CI, 0.96-0.98)., Conclusions: The proposed approach showed high quantitative performance to identify patients with pancreatic neoplasms and to detect MPD dilatation on an independent test cohort. Performance was robust across subgroups of patients with different lesion characteristics and types. Results confirmed the interest to combine a direct lesion detection approach with secondary features such as the MPD diameter, thus indicating a promising avenue for the detection of pancreatic cancer at early stages., Competing Interests: Conflicts of interest and sources of funding: none declared., (Copyright © 2023 Wolters Kluwer Health, Inc. All rights reserved.)

Background: Single photon emission computed tomography (SPECT) myocardial perfusion images (MPI) can be displayed both in traditional short-axis (SA) cardiac planes and polar maps for interpretation and quantification. It is essential to reorient the reconstructed transaxial SPECT MPI into standard SA slices. This study is aimed to develop a deep-learning-based approach for automatic reorientation of MPI., Methods: A total of 254 patients were enrolled, including 226 stress SPECT MPIs and 247 rest SPECT MPIs. Fivefold cross-validation with 180 stress and 201 rest MPIs was used for training and internal validation; the remaining images were used for testing. The rigid transformation parameters (translation and rotation) from manual reorientation were annotated by an experienced nuclear cardiologist and used as the reference standard. A convolutional neural network (CNN) was designed to predict the transformation parameters. Then, the derived transform was applied to the grid generator and sampler in spatial transformer network (STN) to generate the reoriented image. A loss function containing mean absolute errors for translation and mean square errors for rotation was employed. A three-stage optimization strategy was adopted for model optimization: (1) optimize the translation parameters while fixing the rotation parameters; (2) optimize rotation parameters while fixing the translation parameters; (3) optimize both translation and rotation parameters together., Results: In the test set, the Spearman determination coefficients of the translation distances and rotation angles between the model prediction and the reference standard were 0.993 in X axis, 0.992 in Y axis, 0.994 in Z axis, 0.987 along X axis, 0.990 along Y axis and 0.996 along Z axis, respectively. For the 46 stress MPIs in the test set, the Spearman determination coefficients were 0.858 in percentage of profusion defect (PPD) and 0.858 in summed stress score (SSS); for the 46 rest MPIs in the test set, the Spearman determination coefficients were 0.9 in PPD and 0.9 in summed rest score (SRS)., Conclusions: Our deep learning-based LV reorientation method is able to accurately generate the SA images. Technical validations and subsequent evaluations of measured clinical parameters show that it has great promise for clinical use., (© 2023. The Author(s) under exclusive licence to American Society of Nuclear Cardiology.)