A Deep Learning Model for Detecting Rhegmatogenous Retinal Detachment Using Ophthalmologic Ultrasound Images

BackgroundThe preoperative prediction of spread through air spaces (STAS) in patients with early-stage lung adenocarcinoma (LUAD) is crucial for selecting the appropriate surgical approach and improving patient outcomes. Previous research has confirmed that there is a significant correlation between consolidation-to-tumor ratio (CTR) and STAS. This study aimed to develop a Bayesian deep learning (DL) model based on the CTR prior to predict STAS in patients with stage IA LUAD.MethodsThis large-scale diagnostic study included patients with solitary primary invasive LUAD who underwent complete resection between November 2017 and October 2023. Enrolled patients were randomly assigned to training, validation, and test cohorts in a 7:2:1 ratio. Using a variational Bayesian inference framework, we developed a DL model based on the CTR prior (STAS-DLPrior CTR). The performance of STAS-DLPrior CTR was compared with another DL model without the CTR prior (STAS-DLNon-prior CTR) using the receiver operating characteristic (ROC) curve, calibration curve, decision curve analysis (DCA), and clinical impact curve (CIC).ResultsA total of 1,374 patients were included, with 961 in the training cohort, 275 in the validation cohort, and 138 in the test cohort. The results showed that CTR in the STAS-positive group was significantly higher than that in the STAS-negative group [0.63 (interquartile range, 0.36, 0.98) vs. 0.35 (interquartile range, 0.19, 0.60), P<0.001]. Compared to STAS-DLNon-prior CTR, the area under the ROC curve (AUC) tends to be higher for STAS-DLPrior CTR (0.831 vs. 0.731, P=0.06) in the validation cohort, and STAS-DLPrior CTR demonstrated a significantly higher AUC (0.858 vs. 0.637, P=0.008) in the test cohort. Additionally, the calibration curve suggested better calibration for STAS-DLPrior CTR. DCA and CIC also indicated that STAS-DLPrior CTR conferred higher clinical net benefit.ConclusionsThe proposed model based on the CTR prior offers significant advantages in predicting STAS in patients with stage IA LUAD, and incorporating doctors’ knowledge as priors can effectively guide the development of DL models.

Metabolic dysfunction-associated steatotic liver disease (MASLD) can progress to metabolic dysfunction-associated steatohepatitis (MASH) and liver fibrosis, contributing to a heavier global health burden. Non-invasive diagnostic tools developed using machine learning (ML) and deep learning (DL), two representative artificial intelligence algorithms, are increasingly being explored for MASH and its related fibrosis assessment. This study aimed to compare the diagnostic performance of different ML and DL models and identify the top-performing models for diagnosing MASH and associated liver fibrosis. A systematic review and meta-analysis were conducted across PubMed, Web of Science, Embase and Cochrane Library from inception to May 18, 2025. Pooled area under the receiver operator characteristic curve (AUROC) values with 95 % confidence interval (CI) were calculated. Accuracy, specificity, sensitivity, positive predictive values, and negative predictive values were also recorded. Of 4,314 studies initially identified, 106 met the inclusion criteria, with 35 studies (ML: n=28; DL: n=7) providing data for analysis. Logistic Regression and Neural Network are the most commonly algorithms applied in ML and DL, respectively. The pooled AUROCs for diagnosing MASH were 0.833 (95 %CI: 0.806-0.860) for ML models and 0.841 (95 %CI: 0.782-0.900) for DL models. Light Gradient Boosting Machine (LightGBM) and ResNet50 were the best-performing models for diagnosing MASH within ML and DL algorithms, respectively, achieving corresponding AUROCs of 0.920 (95 %CI: 0.916-0.924) and 0.960 (95 %CI: 0.951-0.969). For fibrosis diagnosis, ML models had a pooled AUROC of 0.826 (95 %CI: 0.792-0.860), with Categorical Boosting (CatBoost) achieving the highest AUROC of 0.960 (95 %CI: 0.950-0.970). DL models yielded the pooled AUROC of 0.875 (95 %CI: 0.816-0.934) for fibrosis diagnosis. Both ML and DL models demonstrated strong diagnostic performance for MASH and liver fibrosis, with DL achieving marginally higher AUROCs. AI-driven approaches show promise in MASLD management.

PurposeTo conduct a head-to-head comparison between deep learning (DL) and radiomics models across institutions for predicting microvascular invasion (MVI) in hepatocellular carcinoma (HCC) and to investigate the model robustness and generalizability through rigorous internal and external validation.MethodsThis retrospective study included 2304 preoperative images of 576 HCC lesions from two centers, with MVI status determined by postoperative histopathology. We developed DL and radiomics models for predicting the presence of MVI using B-mode ultrasound, contrast-enhanced ultrasound (CEUS) at the arterial, portal, and delayed phases, and a combined modality (B + CEUS). For radiomics, we constructed models with enlarged vs. original regions of interest (ROIs). A cross-validation approach was performed by training models on one center’s dataset and validating the other, and vice versa. This allowed assessment of the validity of different ultrasound modalities and the cross-center robustness of the models. The optimal model combined with alpha-fetoprotein (AFP) was also validated. The head-to-head comparison was based on the area under the receiver operating characteristic curve (AUC).ResultsThirteen DL models and 25 radiomics models using different ultrasound modalities were constructed and compared. B + CEUS was the optimal modality for both DL and radiomics models. The DL model achieved AUCs of 0.802–0.818 internally and 0.667–0.688 externally across the two centers, whereas radiomics achieved AUCs of 0.749–0.869 internally and 0.646–0.697 externally. The radiomics models showed overall improvement with enlarged ROIs (P < 0.05 for both CEUS and B + CEUS modalities). The DL models showed good cross-institutional robustness (P > 0.05 for all modalities, 1.6–2.1% differences in AUC for the optimal modality), whereas the radiomics models had relatively limited robustness across the two centers (12% drop-off in AUC for the optimal modality). Adding AFP improved the DL models (P < 0.05 externally) and well maintained the robustness, but did not benefit the radiomics model (P > 0.05).ConclusionCross-institutional validation indicated that DL demonstrated better robustness than radiomics for preoperative MVI prediction in patients with HCC, representing a promising solution to non-standardized ultrasound examination procedures.

Area under the curve may hide poor generalisation to external datasets

BackgroundTreatment of sepsis requires prompt initiation of appropriate empiric antibiotics. Coverage for Gram-negative bacilli (GNB) is decided based on risk factors for resistant organisms, as documented in both structured and unstructured data in the electronic health record. We hypothesized that adding large language models (LLMs) that can decode clinical narratives would improve performance of deep learning models in predicting antimicrobial resistance in GNB sepsis.MethodsThis retrospective cohort analysis included all adult patients who met CDC sepsis criteria admitted to 10 BJC hospitals from 1/2018 to 12/2023. Sepsis episodes were classified as community-onset (< 48hrs from hospitalization) or hospital-onset. GNB were stratified as ceftriaxone-susceptible (SS), ceftriaxone-resistant but cefepime-susceptible (RS), and ceftriaxone- and cefepime-resistant (RR). Culture negative or non-GNB sepsis were labeled SS. Deep learning models (XGBoost, DeepAMR) were developed using all available structured data (e.g., demographics, vitals, labs). LLMs (BioGPT, Medical-LLaMa3-8B) were developed on full or partial history and physical exam notes. Models were assessed using area under the receiver operating characteristic curve (AUROC) and area under precision-recall curve (AUPRC).ResultsA total of 29,517 patients were included, for 23,949 community-onset and 25,101 hospital-onset sepsis episodes with RS+RR GNB in 9% and 17%, respectively (Table 1). Relevant features of structured data models included history of prior infections and resistant organisms. DeepAMR was the best performing deep learning model with AUROC 0.85/AUPRC 0.40 and AUROC 0.81/AUPRC 0.50 for community- and hospital-onset sepsis, respectively (Table 2). The best performing LLM was Medical-LLaMa3-8B on full notes (Figure 1), with AUROC 0.74/AUPRC 0.21 and AUROC 0.67/AUPRC 0.30 for community- and hospital-onset sepsis, respectively. Combining deep learning and LLMs did not significantly improve the prediction of resistant infections compared to deep learning alone.ConclusionStructured medical data provided the highest prediction of resistant GNB infections in sepsis. More analysis is needed to understand why the unstructured data did not improve model predictive performance.DisclosuresMaria Cristina Vazquez Guillamet, MD, Bionano: Stocks/Bonds (Public Company)|Charisma: Stocks/Bonds (Public Company)|Ocugen: Stocks/Bonds (Public Company)

To develop and validate a deep learning (DL) model based on quantitative analysis of contrast-enhanced ultrasound (CEUS) images that predicts early recurrence (ER) after thermal ablation (TA) of colorectal cancer liver metastasis (CRLM). Between January 2010 and May 2019, a total of 207 consecutive patients with CRLM with 13,248 slice images at three dynamic phases who received CEUS within 2 weeks before TA were retrospectively enrolled in two centres (153 for the training cohort (TC), 32 for the internal test cohort (ITC), and 22 for the external test cohort (ETC)). Clinical and CEUS data were used to develop and validate the clinical model, DL model, and DL combining with clinical (DL-C) model to predict ER after TA. The performance of these models was compared by the receiver operating characteristic curve (ROC) with the DeLong test. After a median follow-up of 56 months, 49% (99/207) of patients experienced ER. Three key clinical features (preoperative chemotherapy (PC), lymph node metastasis of the primary colorectal cancer (LMPCC), and T stage) were used to develop the clinical model. The DL model yielded better performance than the clinical model in the ETC (AUC: 0.67 for the clinical model, 0.76 for the DL model). The DL-C model significantly outperformed the clinical model and DL model (AUC: 0.78 for the DL-C model in the ETC; both, p < 0.001). The model based on CEUS can achieve satisfactory prediction and assist physicians during the therapeutic decision-making process in clinical practice. • This is an exploratory study in which ablation-related contrast-enhanced ultrasound (CEUS) data from consecutive patients with colorectal cancer liver metastasis (CRLM) were collected simultaneously at multiple institutions. • The deep learning combining with clinical (DL-C) model provided desirable performance for the prediction of early recurrence (ER) after thermal ablation (TA). • The DL-C model based on CEUS provides guidance for TA indication selection and making therapeutic decisions.

We developed a multimodal ultrasound (US) deep learning (DL) fusion model to automatically classify early fibrosis in patients with chronic kidney disease (CKD). This prospective study included patients with CKD who underwent continuous gray-scale US, superb microvascular imaging, and strain elastography from May to November 2022. According to the pathological tubular atrophy and interstitial fibrosis score, patients were divided into minimal and mild groups (affected area ≤10% and 11 − 25% of the total cortical volume, respectively). The dataset was divided into training (70%) and test (30%) sets. A DL model combining the features of the three US modes was developed to predict early fibrosis in patients with CKD. We compared these findings with the area under the receiver operating characteristic curve (AUC) of the clinical model by analyzing the receiver operating characteristic curve in the test set. The AUC of single-mode DL based on gray-scale US, superb microvascular imaging, and strain elastography was 0.682, 0.745, and 0.648, respectively, while that of the multimodal US DL model was 0.86. The accuracy, specificity, and sensitivity of the multimodal US DL model were 0.779, 0.767, and 0.796, respectively, and the negative and positive predictive values were 0.842 and 0.706, respectively. The AUC of the multimodal US DL model was significantly better than that of the single-mode DL and clinical models. The DL algorithm developed using multimodal US images can effectively predict early fibrosis in patients with CKD with significantly greater accuracy than single-mode DL or clinical models.

We aimed at comparing intratumoral and peritumoral deep learning, radiomics, and fusion models in predicting KRAS mutations in rectal cancer using endorectal ultrasound imaging. This study included 304 patients with rectal cancer from Fujian Medical University Union Hospital. The patients were randomly divided into a training group (213 patients) and a test group (91 patients) at a 7:3 ratio. Radiomics and deep learning models were established using primary tumor and peritumoral images. In the optimally performing regions-of-interest, two fusion strategies, a feature-based and a decision-based model, were employed to build the fusion models. The Shapley additive explanation (SHAP) method was used to evaluate the significance of features in the optimal radiomics, deep learning, and fusion models. The performance of each model was assessed using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA). In the test cohort, both the radiomics and deep learning models exhibited optimal performance with a 10-pixel patch extension, yielding AUC values of 0.824 and 0.856, respectively. The feature-based DLRexpand10_FB model attained the highest AUC (0.896) across all study sets. In addition, the DLRexpand10_FB model demonstrated excellent sensitivity, specificity, and DCA. SHAP analysis underscored the deep learning feature (DL_1) as the most significant factor in the hybrid model. The feature-based fusion model DLRexpand10_FB can be employed to predict KRAS gene mutations based on pretreatment endorectal ultrasound images of rectal cancer. The integration of peritumoral regions enhanced the predictive performance of both the radiomics and deep learning models.

Deep learning breast cancer risk models demonstrate improved accuracy compared with traditional risk models but have not been prospectively tested. We compared the accuracy of a deep learning risk score derived from the patient's prior mammogram to traditional risk scores to prospectively identify patients with cancer in a cohort due for screening. We collected data on 119 139 bilateral screening mammograms in 57 617 consecutive patients screened at 5 facilities between September 18, 2017, and February 1, 2021. Patient demographics were retrieved from electronic medical records, cancer outcomes determined through regional tumor registry linkage, and comparisons made across risk models using Wilcoxon and Pearson χ2 2-sided tests. Deep learning, Tyrer-Cuzick, and National Cancer Institute Breast Cancer Risk Assessment Tool (NCI BCRAT) risk models were compared with respect to performance metrics and area under the receiver operating characteristic curves. Cancers detected per thousand patients screened were higher in patients at increased risk by the deep learning model (8.6, 95% confidence interval [CI] = 7.9 to 9.4) compared with Tyrer-Cuzick (4.4, 95% CI = 3.9 to 4.9) and NCI BCRAT (3.8, 95% CI = 3.3 to 4.3) models (P &lt; .001). Area under the receiver operating characteristic curves of the deep learning model (0.68, 95% CI = 0.66 to 0.70) was higher compared with Tyrer-Cuzick (0.57, 95% CI = 0.54 to 0.60) and NCI BCRAT (0.57, 95% CI = 0.54 to 0.60) models. Simulated screening of the top 50th percentile risk by the deep learning model captured statistically significantly more patients with cancer compared with Tyrer-Cuzick and NCI BCRAT models (P &lt; .001). A deep learning model to assess breast cancer risk can support feasible and effective risk-based screening and is superior to traditional models to identify patients destined to develop cancer in large screening cohorts.

This study aims to develop an attention-based deep learning model for distinguishing benign from malignant breast lesions on CESM. Preoperative CESM images of 1239 patients, which were definitely diagnosed on pathology in a multicentre cohort, were divided into training and validation sets, internal and external test sets. The regions of interest of the breast lesions were outlined manually by a senior radiologist. We adopted three conventional convolutional neural networks (CNNs), namely, DenseNet 121, Xception, and ResNet 50, as the backbone architectures and incorporated the convolutional block attention module (CBAM) into them for classification. The performance of the models was analysed in terms of the receiver operating characteristic (ROC) curve, accuracy, the positive predictive value (PPV), the negative predictive value (NPV), the F1 score, the precision recall curve (PRC), and heat maps. The final models were compared with the diagnostic performance of conventional CNNs, radiomics models, and two radiologists with specialised breast imaging experience. The best-performing deep learning model, that is, the CBAM-based Xception, achieved an area under the ROC curve (AUC) of 0.970, a sensitivity of 0.848, a specificity of 1.000, and an accuracy of 0.891 on the external test set, which was higher than those of other CNNs, radiomics models, and radiologists. The PRC and the heat maps also indicated the favourable predictive performance of the attention-based CNN model. The diagnostic performance of two radiologists improved with deep learning assistance. Using an attention-based deep learning model based on CESM images can help to distinguishing benign from malignant breast lesions, and the diagnostic performance of radiologists improved with deep learning assistance.

BackgroundArtificial intelligence–enhanced imaging techniques have demonstrated promising diagnostic potential for carotid plaques, a key cardiovascular and cerebrovascular risk factor. However, previous studies did not systematically synthesize their diagnostic accuracy.ObjectiveThis study aimed to quantitatively explore the diagnostic efficacy of deep learning (DL) and radiomics for extracranial carotid plaques and establish a standardized framework for improving plaque detection.MethodsWe searched the PubMed, Embase, Cochrane, Web of Science, and Institute of Electrical and Electronics Engineers databases to identify studies involving the use of radiomics or DL models to diagnose extracranial carotid artery plaques from inception up to September 24, 2025. The quality of the studies was determined using Quality Assessment of Diagnostic Accuracy Studies for Artificial Intelligence (QUADAS-AI). A meta-analysis was conducted using StataMP (version 17.0; StataCorp) with a bivariate mixed-effects model to calculate pooled sensitivity and specificity, generate summary receiver operating characteristic (SROC) curves, assess Cochran Q statistic and I²-based heterogeneity, and conduct subgroup analyses and regression analysis.ResultsAmong 40 studies comprising 17,246 patients, 34 integrated independent test sets or validation sets in the quantitative statistical analysis. Among them, 24 focused on DL models, 10 on machine learning models based on radiomics. The combined sensitivity, specificity, and area under the SROC curve were 0.88 (95% CI 0.85‐0.91; P<.001; I2=93.58%), 0.89 (95% CI 0.85‐0.92; P<.001; I2=91.38%), and 0.95 (95% CI 0.92‐0.96), respectively. Compared with the machine learning models based on radiomics algorithms, DL models achieved comparable improvements in specificity and area under the SROC curve. It was observed that transfer learning and a large sample size enhanced the diagnostic performance of models. Models used to identify plaque stability and presence had similar diagnostic performances, both of which were more effective in identifying symptomatic plaque models. A total of 7 studies demonstrated that the models that combined clinical features exhibited comparable diagnostic capability to pure DL and radiomics models. Additionally, 7 studies performed external validation, obtaining lower diagnostic performance than in testing groups. Limited regression analysis failed to identify significant sources of heterogeneity, and the limited number of eligible studies restricted more comprehensive subgroup analyses. The high heterogeneity in the study results may be due to different scanning parameters, model architecture, image segmentation, and algorithms.ConclusionsRadiomics algorithms and DL models can effectively diagnose extracranial carotid plaque. However, there are concerns regarding irregularities in research design and the absence of multicenter studies and external validation. Future research should aim to reduce bias risk and enhance the generalizability and clinical orientation of the models.

COVID-19 is highly infectious and causes acute respiratory disease. Machine learning (ML) and deep learning (DL) models are vital in detecting disease from computerized chest tomography (CT) scans. The DL models outperformed the ML models. For COVID-19 detection from CT scan images, DL models are used as end-to-end models. Thus, the performance of the model is evaluated for the quality of the extracted feature and classification accuracy. There are four contributions included in this work. First, this research is motivated by studying the quality of the extracted feature from the DL by feeding these extracted to an ML model. In other words, we proposed comparing the end-to-end DL model performance against the approach of using DL for feature extraction and ML for the classification of COVID-19 CT scan images. Second, we proposed studying the effect of fusing extracted features from image descriptors, e.g., Scale-Invariant Feature Transform (SIFT), with extracted features from DL models. Third, we proposed a new Convolutional Neural Network (CNN) to be trained from scratch and then compared to the deep transfer learning on the same classification problem. Finally, we studied the performance gap between classic ML models against ensemble learning models. The proposed framework is evaluated using a CT dataset, where the obtained results are evaluated using five different metrics The obtained results revealed that using the proposed CNN model is better than using the well-known DL model for the purpose of feature extraction. Moreover, using a DL model for feature extraction and an ML model for the classification task achieved better results in comparison to using an end-to-end DL model for detecting COVID-19 CT scan images. Of note, the accuracy rate of the former method improved by using ensemble learning models instead of the classic ML models. The proposed method achieved the best accuracy rate of 99.39%.

Skin diseases impact millions of people around the world and pose a severe risk to public health. These diseases have a wide range of effects on the skin’s structure, functionality, and appearance. Identifying and predicting skin diseases are laborious processes that require a complete physical examination, a review of the patient’s medical history, and proper laboratory diagnostic testing. Additionally, it necessitates a significant number of histological and clinical characteristics for examination and subsequent treatment. As a disease’s complexity and quantity of features grow, identifying and predicting it becomes more challenging. This research proposes a deep learning (DL) model utilizing transfer learning (TL) to quickly identify skin diseases like chickenpox, measles, and monkeypox. A pre-trained VGG16 is used for transfer learning. The VGG16 can identify and predict diseases more quickly by learning symptom patterns. Images of the skin from the four classes of chickenpox, measles, monkeypox, and normal are included in the dataset. The dataset is separated into training and testing. The experimental results performed on the dataset demonstrate that the VGG16 model can identify and predict skin diseases with 93.29% testing accuracy. However, the VGG16 model does not explain why and how the system operates because deep learning models are black boxes. Deep learning models’ opacity stands in the way of their widespread application in the healthcare sector. In order to make this a valuable system for the health sector, this article employs layer-wise relevance propagation (LRP) to determine the relevance scores of each input. The identified symptoms provide valuable insights that could support timely diagnosis and treatment decisions for skin diseases.

Objective.Functional brain network (FBN) methods are commonly integrated with deep learning (DL) models for EEG analysis. Typically, an FBN is constructed to extract features from EEG data, which are then fed into a DL model for further analysis. Beyond this two-step approach, there is potential to embed FBN construction directly within DL models as a feature extraction module, enabling the models to learn EEG representations end-to-end while incorporating insights from FBNs. However, a critical prerequisite is whether DL models can effectively learn the FBN construction process.Approach.To address this, we propose using DL models to learn FBN matrices derived from EEG data. The ability of DL models to accurately reproduce these matrices would validate their capacity to learn the FBN construction process. This approach is tested on two publicly available EEG datasets, utilizing seven DL models to learn four representative FBN matrices. Model performance is assessed through mean squared error (MSE), Pearson correlation coefficient (Corr), and concordance correlation coefficient (CCC) between predicted and actual matrices.Main results.The results show that DL models achieve low MSE and relatively high Corr and CCC values when learning the Coherence network. Visualizations of predicted and error matrices reveal that while DL models capture the general structure of all four FBNs, certain regions remain difficult to model accurately. Additionally, a pairedt-test comparing global efficiency and nodal degree between predicted and actual networks indicates that most predicted networks significantly differ from the actual networks (p<0.05).Significance.These findings suggest that while DL models can learn the connectivity relationships of certain FBNs, they struggle to capture the intrinsic topological structures. This highlights the irreplaceability of traditional FBN methods in EEG analysis and underscores the need for hybrid strategies that combine FBN methods with DL models for a more comprehensive analysis.

The past few years have seen a surge of applying Deep Learning (DL) models for a wide array of tasks such as image classification, object detection, machine translation, etc. While DL models provide an opportunity to solve otherwise intractable tasks, their adoption relies on them being optimized to meet latency and resource requirements. Benchmarking is a key step in this process but has been hampered in part due to the lack of representative and up-to-date benchmarking suites. This is exacerbated by the fast-evolving pace of DL models. This paper proposes DLBricks, a composable benchmark generation design that reduces the effort of developing, maintaining, and running DL benchmarks on CPUs. DLBricks decomposes DL models into a set of unique runnable networks and constructs the original model's performance using the performance of the generated benchmarks. DLBricks leverages two key observations: DL layers are the performance building blocks of DL models and layers are extensively repeated within and across DL models. Since benchmarks are generated automatically and the benchmarking time is minimized, DLBricks can keep up-to-date with the latest proposed models, relieving the pressure of selecting representative DL models. Moreover, DLBricks allows users to represent proprietary models within benchmark suites. We evaluate DLBricks using $50$ MXNet models spanning $5$ DL tasks on $4$ representative CPU systems. We show that DLBricks provides an accurate performance estimate for the DL models and reduces the benchmarking time across systems (e.g. within $95\%$ accuracy and up to $4.4\times$ benchmarking time speedup on Amazon EC2 c5.xlarge).