Comput Biol Med. 2024 Jun 18;178:108755. doi: 10.1016/j.compbiomed.2024.108755. Online ahead of print.

ABSTRACT

PURPOSE: Impacted teeth are abnormal tooth disorders under the gums or jawbone that cannot take their normal position even though it is time to erupt. This study aims to detect all impacted teeth and to classify impacted third molars according to the Winter method with an artificial intelligence model on panoramic radiographs.

METHODS: In this study, 1197 panoramic radiographs from the dentistry faculty database were collected for all impacted teeth, and 1000 panoramic radiographs were collected for Winter classification. Some pre-processing methods were performed and the images were doubled with data augmentation. Both datasets were randomly divided into 80% training, 10% validation, and 10% testing. After transfer learning and fine-tuning processes, the two datasets were trained with the YOLOv8 deep learning algorithm, a high-performance artificial intelligence model, and the detection of impacted teeth was carried out. The results were evaluated with precision, recall, mAP, and F1-score performance metrics. A graphical user interface was designed for clinical use with the artificial intelligence weights obtained as a result of the training.

RESULTS: For the detection of impacted third molar teeth according to Winter classification, the average precision, average recall, and average F1 score were obtained to be 0.972, 0.967, and 0.969, respectively. For the detection of all impacted teeth, the average precision, average recall, and average F1 score were obtained as 0.991, 0.995, and 0.993, respectively.

CONCLUSION: According to the results, the artificial intelligence-based YOLOv8 deep learning model successfully detected all impacted teeth and the impacted third molar teeth according to the Winter classification system.

PMID:38897151 | DOI:10.1016/j.compbiomed.2024.108755

Comput Biol Med. 2024 Jun 18;178:108733. doi: 10.1016/j.compbiomed.2024.108733. Online ahead of print.

ABSTRACT

BACKGROUND AND OBJECTIVES: Liver segmentation is pivotal for the quantitative analysis of liver cancer. Although current deep learning methods have garnered remarkable achievements for medical image segmentation, they come with high computational costs, significantly limiting their practical application in the medical field. Therefore, the development of an efficient and lightweight liver segmentation model becomes particularly important.

METHODS: In our paper, we propose a real-time, lightweight liver segmentation model named G-MBRMD. Specifically, we employ a Transformer-based complex model as the teacher and a convolution-based lightweight model as the student. By introducing proposed multi-head mapping and boundary reconstruction strategies during the knowledge distillation process, Our method effectively guides the student model to gradually comprehend and master the global boundary processing capabilities of the complex teacher model, significantly enhancing the student model's segmentation performance without adding any computational complexity.

RESULTS: On the LITS dataset, we conducted rigorous comparative and ablation experiments, four key metrics were used for evaluation, including model size, inference speed, Dice coefficient, and HD95. Compared to other methods, our proposed model achieved an average Dice coefficient of 90.14±16.78%, with only 0.6 MB memory and 0.095 s inference speed for a single image on a standard CPU. Importantly, this approach improved the average Dice coefficient of the baseline student model by 1.64% without increasing computational complexity.

CONCLUSION: The results demonstrate that our method successfully realizes the unification of segmentation precision and lightness, and greatly enhances its potential for widespread application in practical settings.

PMID:38897144 | DOI:10.1016/j.compbiomed.2024.108733

Comput Biol Med. 2024 Jun 18;178:108765. doi: 10.1016/j.compbiomed.2024.108765. Online ahead of print.

ABSTRACT

BACKGROUND: Clinical core medical knowledge (CCMK) learning is essential for medical trainees. Adaptive assessment systems can facilitate self-learning, but extracting experts' CCMK is challenging, especially using modern data-driven artificial intelligence (AI) approaches (e.g., deep learning).

OBJECTIVES: This study aims to develop a multi-expert knowledge-aggregated adaptive assessment scheme (MEKAS) using knowledge-based AI approaches to facilitate the learning of CCMK in otolaryngology (CCMK-OTO) and validate its effectiveness through a one-month training program for CCMK-OTO education at a tertiary referral hospital.

METHODS: The MEKAS utilized the repertory grid technique and case-based reasoning to aggregate experts' knowledge to construct a representative CCMK base, thereby enabling adaptive assessment for CCMK-OTO training. The effects of longitudinal training were compared between the experimental group (EG) and the control group (CG). Both groups received a normal training program (routine meeting, outpatient/operation room teaching, and classroom teaching), while EG received MEKAS for self-learning. The EG comprised 22 UPGY trainees (6 postgraduate [PGY] and 16 undergraduate [UGY] trainees) and 8 otolaryngology residents (ENT-R); the CG comprised 24 UPGY trainees (8 PGY and 16 UGY trainees). The training effectiveness was compared through pre- and post-test CCMK-OTO scores, and user experiences were evaluated using a technology acceptance model-based questionnaire.

RESULTS: Both UPGY (z = -3.976, P < 0.001) and ENT-R (z = -2.038, P = 0.042) groups in EG exhibited significant improvements in their CCMK-OTO scores, while UPGY in CG did not (z = -1.204, P = 0.228). The UPGY group in EG also demonstrated a substantial improvement compared to the UPGY group in CG (z = -4.943, P < 0.001). The EG participants were highly satisfied with the MEKAS system concerning self-learning assistance, adaptive testing, perceived satisfaction, intention to use, perceived usefulness, perceived ease of use, and perceived enjoyment, rating it between an overall average of 3.8 and 4.1 out of 5.0 on all scales.

CONCLUSIONS: The MEKAS system facilitates CCMK-OTO learning and provides an efficient knowledge aggregation scheme that can be applied to other medical subjects to efficiently build adaptive assessment systems for CCMK learning. Larger-scale validation across diverse institutions and settings is warranted further to assess MEKAS's scalability, generalizability, and long-term impact.

PMID:38897143 | DOI:10.1016/j.compbiomed.2024.108765

Med Image Anal. 2024 Jun 8;97:103227. doi: 10.1016/j.media.2024.103227. Online ahead of print.

ABSTRACT

Automatic tracking of viral and intracellular structures displayed as spots with varying sizes in fluorescence microscopy images is an important task to quantify cellular processes. We propose a novel probabilistic tracking approach for multiple particle tracking based on multi-detector and multi-scale data fusion as well as Bayesian smoothing. The approach integrates results from multiple detectors using a novel intensity-based covariance intersection method which takes into account information about the image intensities, positions, and uncertainties. The method ensures a consistent estimate of multiple fused particle detections and does not require an optimization step. Our probabilistic tracking approach performs data fusion of detections from classical and deep learning methods as well as exploits single-scale and multi-scale detections. In addition, we use Bayesian smoothing to fuse information of predictions from both past and future time points. We evaluated our approach using image data of the Particle Tracking Challenge and achieved state-of-the-art results or outperformed previous methods. Our method was also assessed on challenging live cell fluorescence microscopy image data of viral and cellular proteins expressed in hepatitis C virus-infected cells and chromatin structures in non-infected cells, acquired at different spatial-temporal resolutions. We found that the proposed approach outperforms existing methods.

PMID:38897031 | DOI:10.1016/j.media.2024.103227

Nature. 2024 Jun 19. doi: 10.1038/s41586-024-07601-y. Online ahead of print.

ABSTRACT

De novo design of complex protein folds using solely computational means remains a substantial challenge1. Here we use a robust deep learning pipeline to design complex folds and soluble analogues of integral membrane proteins. Unique membrane topologies, such as those from G-protein-coupled receptors2, are not found in the soluble proteome, and we demonstrate that their structural features can be recapitulated in solution. Biophysical analyses demonstrate the high thermal stability of the designs, and experimental structures show remarkable design accuracy. The soluble analogues were functionalized with native structural motifs, as a proof of concept for bringing membrane protein functions to the soluble proteome, potentially enabling new approaches in drug discovery. In summary, we have designed complex protein topologies and enriched them with functionalities from membrane proteins, with high experimental success rates, leading to a de facto expansion of the functional soluble fold space.

PMID:38898281 | DOI:10.1038/s41586-024-07601-y

Med Biol Eng Comput. 2024 Jun 20. doi: 10.1007/s11517-024-03150-8. Online ahead of print.

ABSTRACT

Medical image segmentation commonly involves diverse tissue types and structures, including tasks such as blood vessel segmentation and nerve fiber bundle segmentation. Enhancing the continuity of segmentation outcomes represents a pivotal challenge in medical image segmentation, driven by the demands of clinical applications, focusing on disease localization and quantification. In this study, a novel segmentation model is specifically designed for retinal vessel segmentation, leveraging vessel orientation information, boundary constraints, and continuity constraints to improve segmentation accuracy. To achieve this, we cascade U-Net with a long-short-term memory network (LSTM). U-Net is characterized by a small number of parameters and high segmentation efficiency, while LSTM offers a parameter-sharing capability. Additionally, we introduce an orientation information enhancement module inserted into the model's bottom layer to obtain feature maps containing orientation information through an orientation convolution operator. Furthermore, we design a new hybrid loss function that consists of connectivity loss, boundary loss, and cross-entropy loss. Experimental results demonstrate that the model achieves excellent segmentation outcomes across three widely recognized retinal vessel segmentation datasets, CHASE_DB1, DRIVE, and ARIA.

PMID:38898202 | DOI:10.1007/s11517-024-03150-8

Sci Rep. 2024 Jun 19;14(1):14121. doi: 10.1038/s41598-024-64826-7.

ABSTRACT

Sports image classification is a complex undertaking that necessitates the utilization of precise and robust techniques to differentiate between various sports activities. This study introduces a novel approach that combines the deep neural network (DNN) with a modified metaheuristic algorithm known as novel tuna swarm optimization (NTSO) for the purpose of sports image classification. The DNN is a potent technique capable of extracting high-level features from raw images, while the NTSO algorithm optimizes the hyperparameters of the DNN, including the number of layers, neurons, and activation functions. Through the application of NTSO to the DNN, a finely-tuned network is developed, exhibiting exceptional performance in sports image classification. Rigorous experiments have been conducted on an extensive dataset of sports images, and the obtained results have been compared against other state-of-the-art methods, including Attention-based graph convolution-guided third-order hourglass network (AGTH-Net), particle swarm optimization algorithm (PSO), YOLOv5 backbone and SPD-Conv, and Depth Learning (DL). According to a fivefold cross-validation technique, the DNN/NTSO model provided remarkable precision, recall, and F1-score results: 97.665 ± 0.352%, 95.400 ± 0.374%, and 0.8787 ± 0.0031, respectively. Detailed comparisons reveal the DNN/NTSO model's superiority toward various performance metrics, solidifying its standing as a top choice for sports image classification tasks. Based on the practical dataset, the DNN/NTSO model has been successfully evaluated in real-world scenarios, showcasing its resilience and flexibility in various sports categories. Its capacity to uphold precision in dynamic settings, where elements like lighting, backdrop, and motion blur are prominent, highlights its utility. The model's scalability and efficiency in analyzing images from live sports competitions additionally validate its suitability for integration into real-time sports analytics and media platforms. This research not only confirms the theoretical superiority of the DNN/NTSO model but also its pragmatic effectiveness in a wide array of demanding sports image classification assignments.

PMID:38898134 | DOI:10.1038/s41598-024-64826-7

NPJ Precis Oncol. 2024 Jun 19;8(1):134. doi: 10.1038/s41698-024-00623-9.

ABSTRACT

While alterations in nucleus size, shape, and color are ubiquitous in cancer, comprehensive quantification of nuclear morphology across a whole-slide histologic image remains a challenge. Here, we describe the development of a pan-tissue, deep learning-based digital pathology pipeline for exhaustive nucleus detection, segmentation, and classification and the utility of this pipeline for nuclear morphologic biomarker discovery. Manually-collected nucleus annotations were used to train an object detection and segmentation model for identifying nuclei, which was deployed to segment nuclei in H&E-stained slides from the BRCA, LUAD, and PRAD TCGA cohorts. Interpretable features describing the shape, size, color, and texture of each nucleus were extracted from segmented nuclei and compared to measurements of genomic instability, gene expression, and prognosis. The nuclear segmentation and classification model trained herein performed comparably to previously reported models. Features extracted from the model revealed differences sufficient to distinguish between BRCA, LUAD, and PRAD. Furthermore, cancer cell nuclear area was associated with increased aneuploidy score and homologous recombination deficiency. In BRCA, increased fibroblast nuclear area was indicative of poor progression-free and overall survival and was associated with gene expression signatures related to extracellular matrix remodeling and anti-tumor immunity. Thus, we developed a powerful pan-tissue approach for nucleus segmentation and featurization, enabling the construction of predictive models and the identification of features linking nuclear morphology with clinically-relevant prognostic biomarkers across multiple cancer types.

PMID:38898127 | DOI:10.1038/s41698-024-00623-9

Nihon Hoshasen Gijutsu Gakkai Zasshi. 2024 Jun 20. doi: 10.6009/jjrt.2024-1421. Online ahead of print.

ABSTRACT

PURPOSE: To verify the usefulness of a deep learning model for determining the presence or absence of contrast-enhanced myocardium in late gadolinium-enhancement images in cardiac MRI.

METHODS: We used 174 late gadolinium-enhancement myocardial short-axis images obtained from contrast-enhanced cardiac MRI performed using a 3.0T MRI system at the University of Tokyo Hospital. Of these, 144 images were used for training, extracting a region of interest targeting the heart, scaling signal intensity, and data augmentation were performed to obtain 3312 training images. The interpretation report of two cardiology specialists of our hospital was used as the correct label. A learning model was constructed using a convolutional neural network and applied to 30 test data. In all cases, the acquired mean age was 56.4±12.1 years, and the male-to-female ratio was 1 : 0.82.

RESULTS: Before and after data augmentation, sensitivity remained consistent at 93.3%, specificity improved from 0.0% to 100.0%, and accuracy improved from 46.7% to 96.7%.

CONCLUSION: The prediction accuracy of the deep learning model developed in this research is high, suggesting its high usefulness.

PMID:38897968 | DOI:10.6009/jjrt.2024-1421

Int J Surg. 2024 Jun 20. doi: 10.1097/JS9.0000000000001627. Online ahead of print.

ABSTRACT

INTRODUCTION: The postoperative recurrence of gastric cancer has a significant impact on the overall prognosis of patients. Therefore, accurately predicting the postoperative recurrence of gastric cancer is crucial.

METHODS: This retrospective study gathered data from 2,813 gastric cancer patients who underwent radical surgery between 2011 and 2017 at two medical centers. Follow-up was extended until May 2023, and cases were categorized as recurrent or non-recurrent based on postoperative outcomes. Clinical pathological information and imaging data were collected for all patients. A new deep learning signature (DLS) was generated using pretreatment CT images, based on a pre-trained baseline (a customized Resnet50), for predicting postoperative recurrence. The deep learning fusion signature (DLFS) was created by combining the score of DLS with the weighted values of identified clinical features. The predictive performance of the model was evaluated based on discrimination, calibration, and clinical usefulness. Survival curves were plotted to investigate the differences between DLFS and prognosis.

RESULTS: In this study, 2813 patients with gastric cancer (GC) were recruited and allocated into training, internal validation, and external validation cohorts. The DLFS was developed and assessed for its capability in predicting the risk of postoperative recurrence. The DLFS exhibited excellent performance with AUCs of 0.833 (95% CI, 0.809-0.858) in the training set, 0.831 (95% CI, 0.792-0.871) in the internal validation set, and 0.859 (95% CI, 0.806-0.912) in the external validation set, along with satisfactory calibration across all cohorts (P>0.05). Furthermore, the DLFS model significantly outperformed both the clinical model and DLS (P<0.05). High-risk recurrent patients exhibit a significantly poorer prognosis compared to low-risk recurrent patients (P<0.05).

CONCLUSIONS: The integrated model developed in this study, focusing on GC patients undergoing radical surgery, accurately identifies cases at high risk of postoperative recurrence and highlights the potential of DLFS as a prognostic factor for GC patients.

PMID:38896865 | DOI:10.1097/JS9.0000000000001627

Ann Plast Surg. 2024 Jun 18. doi: 10.1097/SAP.0000000000004007. Online ahead of print.

ABSTRACT

BACKGROUND: Hyperpigmentation is a skin disorder characterized by a localized darkening of the skin due to increased melanin production. When patients fail first line topical treatments, secondary treatments such as chemical peels and lasers are offered. However, these interventions are not devoid of risks and are associated with postinflammatory hyperpigmentation. In the quest for novel therapeutic potentials, this study aims to investigate computational methods in the identification of new targeted therapies in the treatment of hyperpigmentation.

METHODS: We used a comprehensive approach, which integrated text mining, interpreting gene lists through enrichment analysis and integration of diverse biological information (GeneCodis), protein-protein association networks and functional enrichment analyses (STRING), and plug-in network centrality parameters (Cytoscape) to pinpoint genes closely associated with hyperpigmentation. Subsequently, analysis of drug-gene interactions to identify potential drugs (Cortellis) was utilized to select drugs targeting these identified genes. Lastly, we used Deep Learning Based Drug Repurposing Toolkit (DeepPurpose) to conduct drug-target interaction predictions to ultimately identify candidate drugs with the most promising binding affinities.

RESULTS: Thirty-four hyperpigmentation-related genes were identified by text mining. Eight key genes were highlighted by utilizing GeneCodis, STRING, Cytoscape, gene enrichment, and protein-protein interaction analysis. Thirty-five drugs targeting hyperpigmentation-associated genes were identified by Cortellis, and 29 drugs, including 16 M2PK1 inhibitors, 11 KRAS inhibitors, and 2 BRAF inhibitors were recommended by DeepPurpose.

CONCLUSIONS: The study highlights the promise of advanced computational methodology for identifying potential treatments for hyperpigmentation.

PMID:38896860 | DOI:10.1097/SAP.0000000000004007

Int J Surg. 2024 Jun 19. doi: 10.1097/JS9.0000000000001808. Online ahead of print.

ABSTRACT

BACKGROUND: Current prognostic models have limited predictive abilities for the growing number of localized (stage I-III) ccRCCs. It is therefore crucial to explore novel preoperative recurrence prediction models to accurately stratify patients and optimize clinical decisions. This purpose of this study was to develop and externally validate a CT-based deep learning (DL) model for pre-surgical disease-free survival (DFS) prediction.

METHODS: Patients with localized ccRCC were retrospectively enrolled from six independent medical centers. Three-dimensional (3D) tumor regions from CT images were utilized as input to architect a ResNet 50 model, which outputted DL computed risk score (DLCR) of each patient for DFS prediction later. The predictive performance of DLCR was assessed and compared to the radiomics model (Rad-Score), clinical model we built and two existing prognostic models (UISS and Leibovich). The complementary value of DLCR to the UISS, Leibovich, as well as Rad-Score were evaluated by stratified analysis.

RESULTS: 707 patients with localized ccRCC were finally enrolled for models' training and validating. The DLCR we established can perfectly stratify patients into low-, intermediate- and high-risks, and outperformed the Rad-Score, clinical model, UISS and Leibovich score in DFS prediction, with a C-index of 0.754 (0.689-0.821) in the external testing set. Furthermore, the DLCR presented excellent risk stratification capacity in subgroups defined by almost all clinic-pathological features. Moreover, patients in the UISS/Leibovich score/Rad-Score stratified low-risk but DLCR-defined intermediate- and high-risk groups were significantly more likely to experience ccRCC recurrences than those of intermediate- and high-risk in DLCR determined low-risk (all Log-rank P values＜0.05).

CONCLUSIONS: Our deep learning model, derived from preoperative CT, is superior to radiomics and current models in precisely DFS predicting of localized ccRCC, and can provide complementary values to them, which may assist more informed clinical decisions and adjuvant therapies adoptions.

PMID:38896853 | DOI:10.1097/JS9.0000000000001808

Med Phys. 2024 Jun 19. doi: 10.1002/mp.17260. Online ahead of print.

ABSTRACT

BACKGROUND: Head and neck (HN) gross tumor volume (GTV) auto-segmentation is challenging due to the morphological complexity and low image contrast of targets. Multi-modality images, including computed tomography (CT) and positron emission tomography (PET), are used in the routine clinic to assist radiation oncologists for accurate GTV delineation. However, the availability of PET imaging may not always be guaranteed.

PURPOSE: To develop a deep learning segmentation framework for automated GTV delineation of HN cancers using a combination of PET/CT images, while addressing the challenge of missing PET data.

METHODS: Two datasets were included for this study: Dataset I: 524 (training) and 359 (testing) oropharyngeal cancer patients from different institutions with their PET/CT pairs provided by the HECKTOR Challenge; Dataset II: 90 HN patients(testing) from a local institution with their planning CT, PET/CT pairs. To handle potentially missing PET images, a model training strategy named the "Blank Channel" method was implemented. To simulate the absence of a PET image, a blank array with the same dimensions as the CT image was generated to meet the dual-channel input requirement of the deep learning model. During the model training process, the model was randomly presented with either a real PET/CT pair or a blank/CT pair. This allowed the model to learn the relationship between the CT image and the corresponding GTV delineation based on available modalities. As a result, our model had the ability to handle flexible inputs during prediction, making it suitable for cases where PET images are missing. To evaluate the performance of our proposed model, we trained it using training patients from Dataset I and tested it with Dataset II. We compared our model (Model 1) with two other models which were trained for specific modality segmentations: Model 2 trained with only CT images, and Model 3 trained with real PET/CT pairs. The performance of the models was evaluated using quantitative metrics, including Dice similarity coefficient (DSC), mean surface distance (MSD), and 95% Hausdorff Distance (HD95). In addition, we evaluated our Model 1 and Model 3 using the 359 test cases in Dataset I.

RESULTS: Our proposed model(Model 1) achieved promising results for GTV auto-segmentation using PET/CT images, with the flexibility of missing PET images. Specifically, when assessed with only CT images in Dataset II, Model 1 achieved DSC of 0.56 ± 0.16, MSD of 3.4 ± 2.1 mm, and HD95 of 13.9 ± 7.6 mm. When the PET images were included, the performance of our model was improved to DSC of 0.62 ± 0.14, MSD of 2.8 ± 1.7 mm, and HD95 of 10.5 ± 6.5 mm. These results are comparable to those achieved by Model 2 and Model 3, illustrating Model 1's effectiveness in utilizing flexible input modalities. Further analysis using the test dataset from Dataset I showed that Model 1 achieved an average DSC of 0.77, surpassing the overall average DSC of 0.72 among all participants in the HECKTOR Challenge.

CONCLUSIONS: We successfully refined a multi-modal segmentation tool for accurate GTV delineation for HN cancer. Our method addressed the issue of missing PET images by allowing flexible data input, thereby providing a practical solution for clinical settings where access to PET imaging may be limited.

PMID:38896829 | DOI:10.1002/mp.17260

IEEE Trans Med Imaging. 2024 Jun 19;PP. doi: 10.1109/TMI.2024.3416744. Online ahead of print.

ABSTRACT

The high burden of lung diseases on healthcare necessitates effective detection methods. Current Computer-aided design (CAD) systems are limited by their focus on specific diseases and computationally demanding deep learning models. To overcome these challenges, we introduce CNN-O-ELMNet, a lightweight classification model designed to efficiently detect various lung diseases, surpassing the limitations of disease-specific CAD systems and the complexity of deep learning models. This model combines a convolutional neural network for deep feature extraction with an optimized extreme learning machine, utilizing the imperialistic competitive algorithm for enhanced predictions. We then evaluated the effectiveness of CNN-O-ELMNet using benchmark datasets for lung diseases: distinguishing pneumothorax vs. non-pneumothorax, tuberculosis vs. normal, and lung cancer vs. healthy cases. Our findings demonstrate that CNN-O-ELMNet significantly outperformed (p < 0.05) state-of-the-art methods in binary classifications for tuberculosis and cancer, achieving accuracies of 97.85% and 97.70%, respectively, while maintaining low computational complexity with only 2481 trainable parameters. We also extended the model to categorize lung disease severity based on Brixia scores. Achieving a 96.20% accuracy in multi-class assessment for mild, moderate, and severe cases, makes it suitable for deployment in lightweight healthcare devices.

PMID:38896522 | DOI:10.1109/TMI.2024.3416744

IEEE Trans Image Process. 2024 Jun 19;PP. doi: 10.1109/TIP.2024.3414269. Online ahead of print.

ABSTRACT

With the increasing availability of cameras in vehicles, obtaining license plate (LP) information via on-board cameras has become feasible in traffic scenarios. LPs play a pivotal role in vehicle identification, making automatic LP detection (ALPD) a crucial area within traffic analysis. Recent advancements in deep learning have spurred a surge of studies in ALPD. However, the computational limitations of on-board devices hinder the performance of real-time ALPD systems for moving vehicles. Therefore, we propose a real-time frame-by-frame LP detector focusing on real-time accurate LP detection. Specifically, video frames are categorized into keyframes and non-keyframes. Keyframes are processed by a deeper network (high-level stream), while non-keyframes are handled by a lightweight network (low-level stream), significantly enhancing efficiency. To achieve accurate detection, we design a knowledge distillation strategy to boost the performance of low-level stream and a feature propagation method to introduce the temporal clues in video LP detection. Our contributions are: (1) A real-time frame-by-frame LP detector for video LP detection is proposed, achieving a competitive performance with popular one-stage LP detectors. (2) A simple feature-based knowledge distillation strategy is introduced to improve the low-level stream performance. (3) A spatial-temporal attention feature propagation method is designed to refine the features from non-keyframes guided by the memory features from keyframes, leveraging the inherent temporal correlation in videos. The ablation studies show the effectiveness of knowledge distillation strategy and feature propagation method.

PMID:38896516 | DOI:10.1109/TIP.2024.3414269

Asian Pac Isl Nurs J. 2024 Jun 19;8:e55321. doi: 10.2196/55321.

ABSTRACT

Artificial intelligence (AI) is reshaping health care, including nursing, across Asia, presenting opportunities to improve patient care and outcomes. This viewpoint presents our perspective and interpretation of the current AI landscape, acknowledging its evolution driven by enhanced processing capabilities, extensive data sets, and refined algorithms. Notable applications in countries such as Singapore, South Korea, Japan, and China showcase the integration of AI-powered technologies such as chatbots, virtual assistants, data mining, and automated risk assessment systems. This paper further explores the transformative impact of AI on nursing education, emphasizing personalized learning, adaptive approaches, and AI-enriched simulation tools, and discusses the opportunities and challenges of these developments. We argue for the harmonious coexistence of traditional nursing values with AI innovations, marking a significant stride toward a promising health care future in Asia.

PMID:38896473 | DOI:10.2196/55321

J Clin Monit Comput. 2024 Jun 19. doi: 10.1007/s10877-024-01179-6. Online ahead of print.

ABSTRACT

Hand hygiene among anesthesia personnel is important to prevent hospital-acquired infections in operating rooms; however, an efficient monitoring system remains elusive. In this study, we leverage a deep learning approach based on operating room videos to detect alcohol-based hand hygiene actions of anesthesia providers. Videos were collected over a period of four months from November, 2018 to February, 2019, at a single operating room. Additional data was simulated and added to it. The proposed algorithm utilized a two-dimensional (2D) and three-dimensional (3D) convolutional neural networks (CNNs), sequentially. First, multi-person of the anesthesia personnel appearing in the target OR video were detected per image frame using the pre-trained 2D CNNs. Following this, each image frame detection of multi-person was linked and transmitted to a 3D CNNs to classify hand hygiene action. Optical flow was calculated and utilized as an additional input modality. Accuracy, sensitivity and specificity were evaluated hand hygiene detection. Evaluations of the binary classification of hand-hygiene actions revealed an accuracy of 0.88, a sensitivity of 0.78, a specificity of 0.93, and an area under the operating curve (AUC) of 0.91. A 3D CNN-based algorithm was developed for the detection of hand hygiene action. The deep learning approach has the potential to be applied in practical clinical scenarios providing continuous surveillance in a cost-effective way.

PMID:38896344 | DOI:10.1007/s10877-024-01179-6

Abdom Radiol (NY). 2024 Jun 19. doi: 10.1007/s00261-024-04301-z. Online ahead of print.

ABSTRACT

PURPOSE: To develop a deep learning (DL) zonal segmentation model of prostate MR from T2-weighted images and evaluate TZ-PSAD for prediction of the presence of csPCa (Gleason score of 7 or higher) compared to PSAD.

METHODS: 1020 patients with a prostate MRI were randomly selected to develop a DL zonal segmentation model. Test dataset included 20 cases in which 2 radiologists manually segmented both the peripheral zone (PZ) and TZ. Pair-wise Dice index was calculated for each zone. For the prediction of csPCa using PSAD and TZ-PSAD, we used 3461 consecutive MRI exams performed in patients without a history of prostate cancer, with pathological confirmation and available PSA values, but not used in the development of the segmentation model as internal test set and 1460 MRI exams from PI-CAI challenge as external test set. PSAD and TZ-PSAD were calculated from the segmentation model output. The area under the receiver operating curve (AUC) was compared between PSAD and TZ-PSAD using univariate and multivariate analysis (adjusts age) with the DeLong test.

RESULTS: Dice scores of the model against two radiologists were 0.87/0.87 and 0.74/0.72 for TZ and PZ, while those between the two radiologists were 0.88 for TZ and 0.75 for PZ. For the prediction of csPCa, the AUCs of TZPSAD were significantly higher than those of PSAD in both internal test set (univariate analysis, 0.75 vs. 0.73, p < 0.001; multivariate analysis, 0.80 vs. 0.78, p < 0.001) and external test set (univariate analysis, 0.76 vs. 0.74, p < 0.001; multivariate analysis, 0.77 vs. 0.75, p < 0.001 in external test set).

CONCLUSION: DL model-derived zonal segmentation facilitates the practical measurement of TZ-PSAD and shows it to be a slightly better predictor of csPCa compared to the conventional PSAD. Use of TZ-PSAD may increase the sensitivity of detecting csPCa by 2-5% for a commonly used specificity level.

PMID:38896250 | DOI:10.1007/s00261-024-04301-z

Eur J Orthop Surg Traumatol. 2024 Jun 19. doi: 10.1007/s00590-024-04032-3. Online ahead of print.

ABSTRACT

PURPOSE: Osteoporosis is a common bone disorder characterized by decreased bone mineral density (BMD) and increased bone fragility, which can lead to fractures and eventually cause morbidity and mortality. It is of great concern that the one-year mortality rate for osteoporotic hip fractures could be as high as 22%, regardless of the treatment. Currently, BMD measurement is the standard method for osteoporosis diagnosis, but it is costly and requires special equipment. While a plain radiograph can be obtained more simply and inexpensively, it is not used for diagnosis. Deep learning technologies had been applied to various medical contexts, yet few to osteoporosis unless they were trained on the advanced investigative images, such as computed tomography. The purpose of this study was to develop a deep learning model using the anteroposterior hip radiograph images and measure its diagnostic accuracy for osteoporosis.

METHODS: We retrospectively collected all anteroposterior hip radiograph images of patients from 2013 to 2021 at a tertiary care hospital. The BMD measurements of the included patients were reviewed, and the radiograph images that had a time interval of more than two years from the measurements were excluded. All images were randomized using a computer-generated unequal allocation into two datasets, i.e., 80% of images were used for the training dataset and the remaining 20% for the test dataset. The T score of BMD obtained from the ipsilateral femoral neck of the same patient closest to the date of the performed radiograph was chosen. The T score cutoff value of - 2.5 was used to diagnose osteoporosis. Five deep learning models were trained on the training dataset, and their diagnostic performances were evaluated using the test dataset. Finally, the best model was determined by the area under the curves (AUC).

RESULTS: A total of 363 anteroposterior hip radiograph images were identified. The average time interval between the performed radiograph and the BMD measurement was 6.6 months. Two-hundred-thirteen images were labeled as non-osteoporosis (T score > - 2.5), and the other 150 images as osteoporosis (T score ≤ - 2.5). The best-selected deep learning model achieved an AUC of 0.91 and accuracy of 0.82.

CONCLUSIONS: This study demonstrates the potential of deep learning for osteoporosis screening using anteroposterior hip radiographs. The results suggest that the deep learning model might potentially be used as a screening tool to find patients at risk for osteoporosis to perform further BMD measurement.

PMID:38896146 | DOI:10.1007/s00590-024-04032-3

J Magn Reson Imaging. 2024 Jun 19. doi: 10.1002/jmri.29476. Online ahead of print.

NO ABSTRACT

PMID:38896101 | DOI:10.1002/jmri.29476