Computer Methods And Programs In Biomedicine[JOURNAL]
Sun
200 papers
Antibiotic resistance prediction with an attention-based bi-LSTM clinical decision support system.
Vouriot L, Rebaudet S, Camiade S … +3 more , Lebsir M, Gaudart J, Urena R
Comput Methods Programs Biomed · 2026 Aug · PMID 42048917 · Publisher ↗
BACKGROUND AND OBJECTIVE: Antimicrobial resistance is recognized by the World Health Organization as a significant global health threat. In clinical practice, the accurate identification of bacterial susceptibility to an...
BACKGROUND AND OBJECTIVE: Antimicrobial resistance is recognized by the World Health Organization as a significant global health threat. In clinical practice, the accurate identification of bacterial susceptibility to antibiotics is crucial. However, clinical laboratories often take several days to complete this process and, in the meantime, physicians rely on probabilistic and empirical reasoning, coupled with local hospital guidelines. METHODS: In this work, we propose an attention-based bidirectional-Long Short-Term Memory recurrent neural network as a clinical decision support system to predict antibiotic resistance at the patient's bedside prior to the arrival of final antimicrobial testing results from the laboratory. More precisely, the model gives predictions at each stage of the bacterial identification process for a set of 47 single antibiotics and combinations, to support clinicians in their prescribing decision. RESULTS: Great results were achieved, with a mean area under the receiver operating characteristic curve and a mean area under the precision-recall curve reaching up to 0.9. The attention mechanism was used to visualize the importance attributed to each feature and to better interpret the prediction results. CONCLUSION: The model has been integrated into a user-friendly and responsive web application, accessible on both mobile phones and desktops, to be used as a prototype clinical decision support system.
Color image security technique combining encryption and data hiding for healthcare applications.
Prasad S, Singh KN, Singh AK … +1 more , Gupta BB
Comput Methods Programs Biomed · 2026 Aug · PMID 42048916 · Publisher ↗
BACKGROUND AND OBJECTIVE: With the rapid growth of internet technologies, the transmission and storage of multimedia data have become increasingly convenient. However, digital images, especially in the healthcare domain,...
BACKGROUND AND OBJECTIVE: With the rapid growth of internet technologies, the transmission and storage of multimedia data have become increasingly convenient. However, digital images, especially in the healthcare domain, are often more popular and sensitive information carriers than other forms of multimedia and plain-text reports. The security of medical information has become a critical issue. METHOD: To address this challenge, we propose a color (in the YCbCr color space) medical image security technique combining encryption and data hiding with usability through thumbnail preservation. Our method introduces four key innovations. First, a novel cascading key is generated using deep learning for image encryption; second, the sensitive chrominance information portion of the medical image is encrypted using a highly secure key, reducing the overall time cost while simultaneously securing the images; third, confidential information is concealed in significant chrominance portion of the color medical images to obtain stego images, substantially preventing potential copyright violations. Finally, thumbnail-preserving encryption (TPE) is applied to the luminance portion to effectively preserve confidentiality and availability. RESULTS: Extensive experiments on two standard medical datasets demonstrate the effectiveness of the proposed method. Our key generation approach achieves a high level of unpredictability and a large key space, along with strong data-hiding and TPE performance with respect to robustness and availability analysis, respectively. CONCLUSION: These features make it particularly suitable for secure healthcare applications.
TrCN-HDC: Enhancing patient security with graphical authentication and cloud-assisted cardiac monitoring.
S G, M V, R S
Comput Methods Programs Biomed · 2026 Aug · PMID 42048915 · Publisher ↗
BACKGROUND AND OBJECTIVE: Heart disease (HD) remains the leading cause of mortality worldwide, emphasizing the need for early detection and accurate diagnosis to improve patient outcomes. The integration of Internet of T...
BACKGROUND AND OBJECTIVE: Heart disease (HD) remains the leading cause of mortality worldwide, emphasizing the need for early detection and accurate diagnosis to improve patient outcomes. The integration of Internet of Things (IoT) technology in healthcare has enabled real-time data collection through smart wearable devices, facilitating continuous monitoring and early identification of HD. However, the low survival rates of sudden heart attacks highlight the necessity for a secure and intelligent patient monitoring system. METHODS: To address this, we propose TrCN-HDC, a Transformer with Capsule Network-based framework designed for IoT-assisted heart disease diagnosis and prediction. The framework ensures patient security and privacy by implementing biometric authentication with the SHA-512 algorithm, securing access to medical data. Patient information is collected through wearable sensors and undergoes pre-processing, including missing value replacement, data normalization, and noise reduction, to enhance data quality. TrCN-HDC framework processes the refined data through Ascended Extensive ResNet (AE-ResNet) for deep feature extraction, followed by Hummingbird Optimization Algorithm (HBOA) for feature selection. The selected features are further refined using the Lightweight Wave Transformer (LW2T) for improved representation before final heart disease classification using Capsule Network (CapNet). RESULTS: The model's performance is evaluated using accuracy, disease prevalence (DP), positive predictive value (PPV), negative predictive value (NPV), sensitivity, specificity, F1-score, and ROC curve analysis. Conclusions- Experimental results demonstrate that TrCN-HDC outperforms existing models, providing a highly accurate, secure, and efficient heart disease detection system in modern smart healthcare environments.
MRI-based two-way fluid-structure interaction simulation for discriminating symptomatic carotid atherosclerosis.
Fu J, Li L, Guo J … +4 more , Zhang X, Zhu C, Chen D, Yu W
Comput Methods Programs Biomed · 2026 Aug · PMID 42044548 · Publisher ↗
OBJECTIVE: To evaluate the additive value of two-way fluid-structure interaction (twFSI)-derived biomechanical metrics for artery-level discrimination of symptomatic carotid disease. METHODS: This single-center retrospec...
OBJECTIVE: To evaluate the additive value of two-way fluid-structure interaction (twFSI)-derived biomechanical metrics for artery-level discrimination of symptomatic carotid disease. METHODS: This single-center retrospective study included 97 patients (125 carotid arteries) who underwent high-resolution vessel wall imaging (HR-VWI). Three-dimensional lumen, vessel wall, and plaque were reconstructed, followed by twFSI simulations to derive hemodynamic and structural indices. Slice-wise morphological descriptors and twFSI-derived biomechanical candidate features were aggregated to the artery level using an attention-based multiple-instance learning framework. Performance was assessed with patient-wise five-fold cross-validation using area under the receiver operating characteristic curve (AUC), accuracy, and macro-averaged F1 score (macro-F1). RESULTS: Symptomatic arteries tended to show lower shear-related exposure and a more disturbed low-shear environment, whereas solid-domain descriptors showed more heterogeneous between-group behavior. A clinical-morphology model (CM) achieved an AUC of 0.716. Incorporating twFSI-derived biomechanical features (CM-BM) improved discrimination (AUC = 0.821), with accuracy increasing from 0.668 to 0.710 and macro-F1 from 0.720 to 0.761. The final CM-BM model retained a concise set of morphological and biomechanical descriptors, including plaque type, plaque angle range, maximum plaque area, maximum total deformation, maximum normalized WSS, and maximum ECAP. CONCLUSIONS: In this single-center cohort, twFSI-derived biomechanical metrics provided incremental discriminative value beyond clinical and morphological variables alone. Attention-based MIL further offered a slice-weighting mechanism that may aid interpretability at the artery level. These findings support the potential of combined structural-dynamic phenotyping for symptom-oriented carotid assessment, while warranting prospective multicenter external validation.
AngioCAD: A public x-ray angiography dataset and an adaptive fusion framework for stenosis detection.
Hosseini MS, Naghsh-Nilchi AR, Safayani M … +4 more , Sadeghi M, Shirvani E, Danesh M, Miramirkhani SA
Comput Methods Programs Biomed · 2026 Aug · PMID 42033972 · Publisher ↗
BACKGROUND AND OBJECTIVE: Coronary artery disease (CAD) is a leading cause of mortality worldwide, underscoring the need for accurate and timely diagnosis. While X-ray coronary angiography remains the clinical gold stand...
BACKGROUND AND OBJECTIVE: Coronary artery disease (CAD) is a leading cause of mortality worldwide, underscoring the need for accurate and timely diagnosis. While X-ray coronary angiography remains the clinical gold standard for detecting stenosis, its manual interpretation is labor-intensive and prone to inter-observer variability. Many existing artificial intelligence-based approaches rely on limited frame-level datasets that lack temporal continuity, artery-specific annotations, and corresponding clinical context, thereby limiting their effectiveness in real-world applications. METHODS: To address these challenges, we introduce AngioCAD, a publicly available dataset comprising angiographic video sequences and structured clinical data from 413 patients. Each case includes detailed stenosis annotations for every coronary artery, such as 100% stenosis in the proximal segment of the right coronary artery, along with demographic and laboratory information. This dataset supports a broad range of CAD-related tasks, including diagnosis, view classification, and stenosis detection, through both image- and attribute-based analysis. We further propose a deep learning framework for stenosis detection based on video modeling. The model integrates extracted features from two convolutional neural networks via an adaptive fusion module that learns attention weights (α) to prioritize the most informative feature stream for each case. RESULTS: The framework achieves superior performance across multiple evaluation metrics, including F1-score and PR-AUC. Furthermore, we show that incorporating discretized and normalized clinical attributes improves classification performance in classical models, with a polynomial-kernel SVM achieving an F1-score of 89.78%. CONCLUSIONS: These findings highlight the potential of the AngioCAD dataset and adaptive video modeling for improving automated CAD and stenosis detection.
Multimodal neural operators for real-time biomechanical modelling of traumatic brain injury.
Agarwal A, Sarkar DR, Goswami S
Comput Methods Programs Biomed · 2026 Aug · PMID 42033971 · Publisher ↗
BACKGROUND: Traumatic brain injury (TBI) remains a major public health concern, with over 69 million cases annually worldwide. Accurate patient-specific biomechanical modeling is critical for injury risk assessment, but...
BACKGROUND: Traumatic brain injury (TBI) remains a major public health concern, with over 69 million cases annually worldwide. Accurate patient-specific biomechanical modeling is critical for injury risk assessment, but it requires integrating heterogeneous data sources such as volumetric neuroimaging, scalar demographic parameters, and acquisition metadata. Conventional finite element solvers can perform this modeling, yet they remain too computationally expensive for time-sensitive clinical settings. Neural operators have emerged as a promising alternative by learning resolution-invariant mappings between function spaces at orders-of-magnitude faster inference. However, while recent work has introduced parameter-conditioned and multi-input operator architectures for incorporating auxiliary scalar or geometric inputs, the systematic integration of volumetric medical imaging with heterogeneous scalar metadata within an operator learning framework remains underexplored, particularly for biomechanical prediction tasks involving patient-specific anatomical and demographic variability. OBJECTIVE: This study presents a systematic investigation of multimodal neural operator architectures for brain biomechanics, evaluating strategies for fusing heterogeneous input modalities such as volumetric anatomical imaging, scalar demographic features, and acquisition parameters to predict full-field brain displacement fields from MRE data. METHODS: We reformulated TBI modeling as a multimodal operator learning problem and proposed two fusion strategies: field projection for Fourier Neural Operator (FNO) based architectures (broadcasting scalars onto spatial grids) and branch decomposition for Deep Operator Networks (DeepONet) (separate encoding with multiplicative fusion). Four architectures (FNO, Factorized FNO (F-FNO), Multi-Grid FNO (MG-FNO), DeepONet) were extended with multimodal fusion mechanisms and evaluated on 249 in vivo Magnetic Resonance Elastography (MRE) datasets across physiologically relevant frequencies (20 to 90 Hz). RESULTS: DeepONet achieved the highest accuracy on real displacement fields (MSE = 0.0039, 90.0% accuracy) with the fastest inference (3.83 it/s) and fewest parameters (2.09M), while MG-FNO achieved the best performance on imaginary fields (MSE = 0.0058, 88.3% accuracy) with the lowest GPU memory among FNO variants (7.12 GB). No single architecture dominated across all criteria, revealing distinct trade-offs between accuracy, spatial fidelity, and computational cost. CONCLUSION: The results demonstrate that neural operators augmented with multimodal fusion mechanisms can accurately predict full-field brain displacement from heterogeneous biomedical inputs, with inference times orders of magnitude faster than finite element solvers. The systematic comparison of fusion strategies and architectures provides practical guidance for selecting operator learning approaches in biomedical settings where heterogeneous data integration is required.
CrohnTwin-X: An explainable digital twin framework for predicting postoperative recurrence in Crohn's disease using multi-omics and clinical features.
Liu Z, Zhou D, Yang Z
Comput Methods Programs Biomed · 2026 Aug · PMID 42030700 · Publisher ↗
BACKGROUND: Crohn's disease (CD) is a chronic inflammatory disorder with a high rate of postoperative recurrence, posing major challenges for long-term management. Conventional clinical indices have limited capacity for...
BACKGROUND: Crohn's disease (CD) is a chronic inflammatory disorder with a high rate of postoperative recurrence, posing major challenges for long-term management. Conventional clinical indices have limited capacity for individualized prognostic prediction. While Digital Twin (DT) technology-virtual replicas of individual patients-offers a transformative paradigm for clinical decision support, its application in CD postoperative management remains largely unexplored. METHODS: We developed CrohnTwin-X, an explainable digital twin framework that integrates transcriptomic and clinical features to predict postoperative recurrence in CD. Transcriptome and clinical data from GSE186582 (n = 464) and GSE102133 (n = 33) were obtained from GEO. After normalization and batch correction, KEGG pathway activation was quantified using gene set variation analysis and used for functional pathway-driven digital phenotyping to cluster patients into distinct digital-twin subtypes. Differentially expressed genes across subtypes were treated as key feature encoders and modeled using least absolute shrinkage and selection operator and multivariable logistic regression to derive a 17-gene biological fingerprint and twin inference engine. A clinician-digital twin interface (nomogram) was constructed by combining the gene-based risk score with clinical variables. Model performance was assessed by receiver operating characteristic curves, area under the curve (AUC), calibration, and decision curve analysis. RESULTS: CrohnTwin-X stratified patients into three pathway-defined digital phenotypes with distinct immune-metabolic features and postoperative outcomes. The 17-gene fingerprint showed robust performance with AUC values of 0.876 (95% CI: 0.830-0.921), 0.714 (95% CI: 0.640-0.788), 0.789 (95% CI: 0.743-0.835), and 0.800 (95% CI: 0.650-0.950) in the training, testing, total, and external validation sets, respectively. CONCLUSIONS: CrohnTwin-X is a novel, explainable digital twin framework for predicting postoperative recurrence in CD by integrating multi-omics and clinical features, bridging gastroenterology, data science, and biomedical engineering, and supporting personalized postoperative surveillance and treatment decision-making.
Few-shot learning for surgical phase recognition: Performance and generalization in cholecystectomy.
Bajraktari F, Asmußen R, Giacoppo GA … +1 more , Pott PP
Comput Methods Programs Biomed · 2026 Aug · PMID 42030699 · Publisher ↗
BACKGROUND AND OBJECTIVE: The automated recognition of surgical phases in intraoperative videos represents a critical milestone in the digital transformation of surgery. It forms the foundation for surgical assistance sy...
BACKGROUND AND OBJECTIVE: The automated recognition of surgical phases in intraoperative videos represents a critical milestone in the digital transformation of surgery. It forms the foundation for surgical assistance systems and enhanced decision support, contributing to increased safety, efficiency, and precision in surgical procedures. Traditional deep learning methods often fall short in this domain due to their dependency on extensive annotated datasets, which are challenging to obtain in medical contexts due to privacy concerns and data scarcity. This study explores the potential of few-shot learning as a paradigm for overcoming data limitations in surgical phase recognition. METHODS: By leveraging the ability to generalize from minimal examples, a transformer-based few-shot learning (FSL) model for action recognition was adapted to the recognition of surgical phases using the Cholec80 dataset, which consists of videos from cholecystectomy surgeries. The model's performance was evaluated across three experimental splits to assess domain-specific and cross-domain performance: Split 1, where the model was trained on a surgical dataset and tested on Cholec80; Split 2, which introduced variations in surgical environments; and Split 3, where the model was trained on action recognition data and tested on surgical data. RESULTS: The model achieved test accuracies of 89.0%, 75.4%, and 49.1% in these splits, respectively. While FSL demonstrates strong applicability to surgical data, domain-specific training remains crucial for optimal performance. Notably, these results were obtained using only a few labeled support examples per phase, illustrating the data efficiency of the approach. CONCLUSION: This study provides an initial foundation for applying few-shot learning to surgical phase recognition and demonstrates its feasibility under low-label and transfer settings. While domain-specific training remains important, the results indicate that FSL is a promising direction for surgical workflow analysis when annotation resources are limited.
Automation is not yet credibility: on cohort realism, uncertainty propagation, and context-of-use validation in cardiac in silico trials.
Madrid M, Argueta E, Zablah I
Comput Methods Programs Biomed · 2026 Aug · PMID 42030698 · Publisher ↗
Abstract loading — click title to view on PubMed.
Synthesis of coronary 4D CT Image by denoising diffusion probabilistic model.
Han TH, Kim YW, Lee HJ … +8 more , Kim JS, Lee SG, Yang DH, Oh HM, Kim D, Shin SY, Song S, Lee JS
Comput Methods Programs Biomed · 2026 Aug · PMID 42025230 · Publisher ↗
PURPOSE: Fluctuations in the pressure drop during the cardiac cycle can provide prognostic information for coronary artery disease (CAD). However, 4D computed tomography (CT) is required for time-variant flow analysis, w...
PURPOSE: Fluctuations in the pressure drop during the cardiac cycle can provide prognostic information for coronary artery disease (CAD). However, 4D computed tomography (CT) is required for time-variant flow analysis, which results in high doses of radiation exposure. In this study, we propose a novel diffusion-based framework for synthesizing physiologically consistent 4D CT images and performing 4D CT flow analysis. METHODS: A denoising diffusion probabilistic model (DDPM) integrated with a deformation module was used for precise anatomical reconstruction. Subsequently, a computational fluid dynamics (CFD) model coupled with quasi-steady fluid-structure interaction (FSI) was utilized to calculate the 4D hemodynamic flow field. RESULTS: The model achieved a peak signal-to-noise ratio of 32.01 and a structural similarity index measure of 0.937. After 3D construction and segmentation, the average Dice coefficient was 0.973. Furthermore, the computational fluid analysis was also performed with a fractional flow reserve (FFR) accuracy of 90.5%, demonstrating its efficacy in reducing radiation exposure without compromising diagnostic quality. CONCLUSION: Our results demonstrate that this synthesized 4D CT-based hemodynamic approach provides time-variant information for CAD diagnosis. This method offers valuable guidance for clinical decision-making as well as the possibility of prognostic information based on dynamic lumen evaluation.
Uncertainty-aware AI for tumor subtyping with histology and immunohistochemistry: A multi-center study in Renal Cell Carcinoma.
Hosseini SMM, Hannetel P, Di Cataldo S … +5 more , Descombes X, Sibony M, Decaussin M, Ponzio F, Ambrosetti D
Comput Methods Programs Biomed · 2026 Aug · PMID 42019354 · Publisher ↗
BACKGROUND AND OBJECTIVE: Accurate classification of Renal Cell Carcinoma (RCC) subtypes is essential for personalized therapy, as prognostic outcomes and treatment responses differ markedly among subtypes. Most artifici...
BACKGROUND AND OBJECTIVE: Accurate classification of Renal Cell Carcinoma (RCC) subtypes is essential for personalized therapy, as prognostic outcomes and treatment responses differ markedly among subtypes. Most artificial intelligence applications in RCC focus on histology, while immunohistochemistry applications mainly quantify biomarkers rather than perform subtype classification. In clinical diagnostics, immunohistochemistry remains a valuable complement to histology, though its implementation is limited by cost, labor intensity, and resource constraints. This study aims to develop an uncertainty-aware artificial intelligence framework that integrates histological and immunohistochemistry data to improve RCC subtype classification and optimize laboratory workflows. METHODS: We designed a hierarchical, pathologist-guided artificial intelligence framework that integrates uncertainty estimation into RCC subtype classification. High-confidence histological predictions produced by deep learning models are accepted directly, while low-confidence cases automatically trigger targeted immunohistochemistry analysis automatically analyzed using dedicated machine learning algorithms. To address staining variability across institutions, a CycleGAN-based stain transfer module was employed to harmonize color domains and enhance generalizability. The framework was evaluated on multi-center datasets encompassing different staining protocols. RESULTS: The proposed integrated framework demonstrated significant diagnostic performance improvements. Patient-level accuracy reached 97.5% on internal cross-validation and 95% on external cohorts when selective immunohistochemistry refinement was applied. The uncertainty-driven immunohistochemistry module reduced redundant biomarker testing to approximately one-fourth of all cases while maintaining or improving classification confidence. The stain transfer module also effectively mitigated inter-laboratory color discrepancies, supporting consistent model performance across different centers. CONCLUSIONS: Our framework combines histology and immunohistochemistry to deliver accurate and efficient subtype classification. By selectively activating immunohistochemistry analysis for low-confidence predictions on histology, the approach optimizes biomarker use and laboratory resources while maintaining high diagnostic reliability. The study highlights the potential of combining deep learning-based histology with targeted immunohistochemistry biomarker assessment to advance precision and reproducibility in cancer diagnostics workflows.
Automatic selection of optical coherence tomography images for prognostic prediction models in age-related macular degeneration.
Deng Q, Kishimoto K, Sugiyama O … +2 more , Miyake M, Tamura H
Comput Methods Programs Biomed · 2026 Aug · PMID 42019353 · Publisher ↗
BACKGROUND AND OBJECTIVE: Age-related macular degeneration (AMD) is a leading cause of blindness. Current standard treatments require frequent intravitreal injections and entail high costs, placing a heavy burden on pati...
BACKGROUND AND OBJECTIVE: Age-related macular degeneration (AMD) is a leading cause of blindness. Current standard treatments require frequent intravitreal injections and entail high costs, placing a heavy burden on patients and healthcare providers. These challenges often lead to treatment discontinuation or overtreatment, highlighting the need for personalized AMD management. Accurate early prediction of long-term treatment outcomes is critical for optimizing these strategies. However, most existing prognostic models rely heavily on manual image selection by ophthalmologists. This labor-intensive process, which requires carefully selecting suitable images from large volumes of electronic medical record (EMR) data, significantly hinders real-world implementation. This study proposes an automated deep learning framework to select appropriate images from extensive EMR-stored optical coherence tomography (OCT) reports, thereby reducing the reliance on manual curation. METHODS: We developed a vision transformer (ViT)-based architecture to perform automated image selection. The model integrates features from the fundus infrared reflectance (IR) images and the corresponding OCT images presented in the clinical reports. RESULTS: Compared to related works using a single OCT image input, the proposed method outperformed to the baseline. At training the pretrained ViT framework achieved an overall accuracy of 89% in identifying suitable images. Furthermore, using the images selected by our method improved the downstream prognostic prediction accuracy by an average of 3.1 percentage points. Confidence scores showed a statistically significant difference between the proposed method and the baseline (p = 0.026). The 95% confidence interval (CI) of the performance difference was [0.002, 0.066]. CONCLUSIONS: These results demonstrate the feasibility of an automated module capable of reliably identifying suitable images for AMD prognosis. By streamlining image selection workflows, this approach can enhance clinical efficiency, support the accurate estimation of long-term treatment effects, and facilitate treatment planning.
Exploring three-dimensional reconstruction with Neural Radiance Field (NeRF) for coronary roadmap navigation and view-planning in X-ray coronary angiography: A feasibility study.
Vermeer AJE, Maas KWH, McCaughan VM … +7 more , Ruijters D, Pezzotti N, Vlaar PJ, Tonino PAL, van de Vosse FN, Vilanova A, van 't Veer M
Comput Methods Programs Biomed · 2026 Jul · PMID 42019103 · Publisher ↗
BACKGROUND AND OBJECTIVES: Three-dimensional (3D) reconstruction from X-ray coronary angiograms could enhance diagnosis and guide treatment of coronary artery disease. This study investigates the feasibility of applying...
BACKGROUND AND OBJECTIVES: Three-dimensional (3D) reconstruction from X-ray coronary angiograms could enhance diagnosis and guide treatment of coronary artery disease. This study investigates the feasibility of applying Neural Radiance Field (NeRF), a deep learning technique capable of automatic 3D reconstruction from a few views, to two clinical applications: (1) generating a coronary overlay ("roadmap") to assist navigation during interventions without contrast administration, and (2) predicting optimal viewing angles to support procedural planning. We will gather feedback by involving end users in the early evaluation of our approach, providing insights into its clinical relevance. METHODS: The 3D coronary tree was reconstructed using NeRF from various combinations of segmented angiographic views. The resulting 3D reconstructions were re-projected to the original viewing angles to generate 2D model-derived images. These were evaluated by four reviewers using a qualitative questionnaire focused on image quality, coronary topology, and visual clutter. RESULTS: Over 89% of NeRF-based roadmaps were rated as at least acceptable. In contrast, fewer than one-third of the predicted views for view-planning were considered minimally acceptable. Assessment varied between reviewers, particularly in scoring coronary topology and visual clutter, though these differences did not fully account for the variability in overall rated quality. CONCLUSION: NeRF-based 3D reconstruction from X-ray coronary angiography was adequate for generating coronary roadmaps, but inadequate for view-planning. Further improvements are needed in reconstructing accurate 3D coronary topology, especially in the robustness of the model under few angiographic projections, before clinical adoption of NeRF is feasible.
Echo-SMADS: A hierarchical planning model for predicting ejection fraction using echocardiography.
Zhou Y, Tian J, Kang M … +1 more , Kang K
Comput Methods Programs Biomed · 2026 Jul · PMID 42019102 · Publisher ↗
BACKGROUND AND OBJECTIVE: Current deep learning approaches for predicting ejection fraction primarily rely on end-to-end regression. While effective in certain cases, these methods often lack intermediate structural foun...
BACKGROUND AND OBJECTIVE: Current deep learning approaches for predicting ejection fraction primarily rely on end-to-end regression. While effective in certain cases, these methods often lack intermediate structural foundations and limit the interpretability of the clinical decision-making process. This limitation reduces transparency and may hinder clinical adoption. The objective of this study is to design a clinically aligned, modular system that emulates the diagnostic workflow of physicians, improving interpretability, stability, and applicability in real-world ejection fraction assessment. METHODS: We propose Echo-SMADS, a clinically aligned system that adopts the concept of hierarchical planning from artificial intelligence. The overall prediction task is decomposed into three clinically relevant subtasks: structure identification, phase selection, and volume estimation. These subtasks are implemented as decoupled functional modules, each optimized independently while producing interpretable intermediate outputs. The modular design allows reasoning evidence to propagate across stages, enabling a transparent diagnostic reasoning path. RESULTS: Experiments were conducted on the EchoNet-Dynamic dataset. Echo-SMADS achieved a mean absolute error of 5.48 ± 0.17 and a root mean square error of 7.64 ± 0.20 for ejection fraction prediction. The results demonstrate improved performance stability compared with traditional end-to-end models, while providing meaningful intermediate outputs that enhance interpretability and trustworthiness. CONCLUSIONS: Echo-SMADS integrates modular, clinically aligned components into a structured diagnostic process that closely reflects real-world clinical workflows. By combining interpretability, physical grounding, and performance stability, the proposed system offers a promising approach for reliable ejection fraction prediction, with potential for future clinical application.
MUSIOMICS: A multi-region radiomics framework that outperforms single-region analysis in classifying malignant pulmonary nodules.
Tu SJ, Tran VT, Wu CT
Comput Methods Programs Biomed · 2026 Jul · PMID 42013641 · Publisher ↗
BACKGROUND AND OBJECTIVE: Radiomic studies in lung cancer have primarily analyzed single-domain features extracted separately from intranodular (Zone-1) or perinodular (Zone-2) regions, potentially overlooking their biol...
BACKGROUND AND OBJECTIVE: Radiomic studies in lung cancer have primarily analyzed single-domain features extracted separately from intranodular (Zone-1) or perinodular (Zone-2) regions, potentially overlooking their biological interdependence. We developed MUSIOMICS (Multiregional Unified and Spatially Integrated Oncologic Model for Imaging-based Connected Structures), a multi-region radiomic framework, and constructed two region-dependent delta-radiomic models (Delta-1 and Delta-2). Their performance was evaluated and validated in classifying primary versus metastatic pulmonary nodules. METHODS: A total of 443 malignant pulmonary nodules (training set, n = 360; test set, n = 83) were retrospectively analyzed. Zone-1 and Zone-2 vol were delineated using LIFEx software. The MUSIOMICS framework was applied to construct two spatial delta-radiomic models that extracted features from both zones of different biological roles (e.g. Zone-1 and Zone-2). These features of different strengths were then fused into a single effective delta-feature. Predictive models were developed in the training dataset using a two-stage feature selection strategy and three classifiers (Random Forest, AdaBoost, and Support Vector Machine [SVM]). A two-sample t-test was applied to both the training and independent test datasets to identify reproducible statistically significant (RSS) delta-features. SHapley Additive exPlanations (SHAP) analysis was performed to rank feature importance and identify informative delta-features. RSS and informative delta-features together were combined to characterize the spatial delta-radiomic models. RESULTS: In the independent test dataset, spatial delta-radiomic models (Delta-1: 82%, Delta-2: 81%) outperformed single-region models (Zone-1: 75%, Zone-2: 67%), producing a 6-15% improvement in predictive accuracy. Among all classifiers, Delta-1 combined with SVM achieved the highest performance (accuracy, 86%; area under receiver operating characteristic curve, 0.90). The t-test identified 58 and 48 RSS delta-features for Delta-1 and Delta-2, respectively, with 40 overlapping across both models. Among classifiers, GLCM_DV and Intensity_QCD were consistently top-ranked contributors in Delta-1, with Intensity_QCD identified as the most informative feature in both Delta-1 and Delta-2. CONCLUSIONS: Spatial delta-radiomics based on MUSIOMICS integrates complementary information from biologically connected intranodular and perinodular compartments, achieving higher and more reproducible predictive performance than conventional single-zone models for characterizing malignant pulmonary nodules.
Enhancing breast mass detection: Super-resolution multi-spectral transmission imaging with unstructured sinusoidal illumination.
Liu F, Zhou X, Wang J
Comput Methods Programs Biomed · 2026 Jul · PMID 42013640 · Publisher ↗
BACKGROUND AND OBJECTIVE: Optical transmission imaging of biological tissues is often hindered by blurring caused by light absorption and scattering, limiting its accuracy in detecting early-stage abnormalities such as b...
BACKGROUND AND OBJECTIVE: Optical transmission imaging of biological tissues is often hindered by blurring caused by light absorption and scattering, limiting its accuracy in detecting early-stage abnormalities such as breast tumors. This study aims to enhance the resolution and quality of multi-spectral transmission images by integrating unstructured sinusoidal illumination with super-resolution reconstruction techniques. METHODS: Multi-spectral transmission images of tissue-mimicking phantoms were acquired at four wavelengths (435 nm, 546 nm, 700 nm, and 860 nm) under 3.5 Hz unstructured sinusoidal illumination. Images were extracted using Fast Fourier Transform (FFT) and combined into pseudo-color images. Six super-resolution methods-Locally Linear Embedding (LLE), Sparse Coding Super-Resolution (ScSR), Anchored Neighborhood Regression Super-Resolution (ANRSR), Super-Resolution Convolutional Neural Network (SRCNN), Fast Super-Resolution Convolutional Neural Network (FSRCNN), and Efficient Sub-Pixel Convolutional Neural Network (ESPCN)-were applied to enhance image quality. The enhanced images were evaluated using full-reference image quality assessment (FR-IQA) metrics, and heterogeneity detection accuracy was validated using the Faster-RCNN model. RESULTS: The proposed combination significantly improved image quality across all metrics. Among the methods, LLE achieved the highest detection accuracy, with mean Average Precision (mAP) values of 97.57%, 98.26%, and 99.17% for detecting two, four, and seven types of heterogeneities, respectively. LLE also outperformed other methods in most FR-IQA metrics. CONCLUSIONS: Integrating unstructured sinusoidal illumination with super-resolution reconstruction, particularly the LLE method, effectively produces high-quality multi-spectral transmission images. This approach holds strong potential for improving early breast cancer screening accuracy in clinical settings.
A modular deep learning pipeline for stromal TILs scoring in breast cancer H&E slides.
Abdelazeez S, Ahmed F, Adalid L … +5 more , Siemion K, Lopez C, Lejeune M, Rashwan H, Korzynska A
Comput Methods Programs Biomed · 2026 Aug · PMID 42013559 · Publisher ↗
BACKGROUND AND OBJECTIVE: Tumor-infiltrating lymphocytes (TILs) are an important indicator of immune activity in breast cancer, yet scoring them consistently on H&E slides remains challenging in routine pathology. This w...
BACKGROUND AND OBJECTIVE: Tumor-infiltrating lymphocytes (TILs) are an important indicator of immune activity in breast cancer, yet scoring them consistently on H&E slides remains challenging in routine pathology. This work presents a modular deep learning pipeline that delivers fully automated and continuous stromal TILs (sTILs) scores in line with the International Immuno-Oncology Biomarker Working Group (IIOBWG) guidelines. METHODS: The pipeline combines three components: a TIL segmentation model refined through pathologist-guided active learning, a robust stroma segmentation network based on an enhanced DeepLabV3+, and a lightweight regression module that learns how TILs distribute within stromal regions. A new adaptive aggregation strategy integrates patch-level predictions into a single, clinically meaningful score while accounting for heterogeneous infiltration. RESULTS: The system was evaluated on two independent datasets (60 and 112 WSIs) with expert-annotated ROIs, achieving strong agreement with pathologists (Pearson of 0.814; ICC of 0.808). CONCLUSIONS: Importantly, the pipeline is interpretable: each stage produces human-readable outputs (stroma masks, TIL-in-stroma maps), and SegGradCAM visualizations confirm that predictions rely on biologically relevant tissue regions. These findings demonstrate the pipeline's potential as a reliable and clinically adaptable tool for standardized, fully automated TILs quantification in breast cancer pathology. The source code and pretrained models are publicly available at https://github.com/Shrief-Abdelazeez/TILs-Scoring.
In silico modeling of transcatheter heart valve oversizing and ellipticity, Part II: Effects on leaflet mechanics, hemodynamics, and stent deflection contributing to thrombogenic risk and structural degeneration.
Boxwell S, Cahalane RME, Armfield D … +5 more , Hickey W, Cook S, Kelly P, Cardiff P, McNamara LM
Comput Methods Programs Biomed · 2026 Jul · PMID 42001655 · Publisher ↗
BACKGROUND AND OBJECTIVES: Transcatheter aortic valve implantation (TAVI) is the leading treatment for aortic stenosis. Self-expanding transcatheter heart valves (THVs) are oversized to prevent paravalvular leakage and t...
BACKGROUND AND OBJECTIVES: Transcatheter aortic valve implantation (TAVI) is the leading treatment for aortic stenosis. Self-expanding transcatheter heart valves (THVs) are oversized to prevent paravalvular leakage and then deployed over the diseased native valve. However, this can result in incomplete expansion and elliptical deployment, which may influence thrombogenic risk and structural degeneration, although this is not fully understood. METHODS: In this study, we utilized a validated in silico framework to assess the impact of THV oversizing and ellipticity on leaflet mechanics, hemodynamic shear stress and stent deformation, which are indicators of structural degeneration and thrombogenicity. We simulated self-expansion of a deformable THV stent within an idealized aortic annulus, applied pulsatile loading conditions representative of the cardiac cycle and then evaluated post-deployment frame deformation, leaflet mechanics, hemodynamics and stent fatigue. RESULTS: We predicted stent-frame decoupling of the supra-annular THV, with increased expansion and circularity at the functional valve level compared to the inflow. THV oversizing reduced valve expansion at the supra-annular valve level (< 90% expansion), which increased leaflet coaptation and pinwheeling, but reduced peak leaflet stresses and stent deflection compared to nominal sizing. Oversizing also altered hemodynamics, causing early mainstream flow separation, which increased leaflet oscillatory shear and viscous shear stress downstream of the THV, potentially increasing thrombogenic risk and promoting tissue degeneration. THV ellipticity induced heterogenous stent deflections, leading to variable leaflet stress distributions and coaptation mismatch. CONCLUSION: We propose that flexible THV stents may mitigate adverse effects of elliptical deployment and emphasize the importance of assessing THV expansion through fluoroscopy and considering post-TAVI balloon-dilatation to increase expansion and improve long-term functional valve performance.
Radiomics Applicability Domain Analysis Classification Framework (RADAN-CF): A method for evaluating prediction reliability in radiomics.
Rodríguez-Belenguer P, Marfil-Trujillo M, Vraka A … +7 more , Tsiknakis M, Papanikolaou N, Regge D, Marias K, Cerdá-Alberich L, Martí-Bonmatí L, ProCAncer-I Consortium
Comput Methods Programs Biomed · 2026 Jul · PMID 42001654 · Publisher ↗
BACKGROUND AND OBJECTIVE: Radiomics-based machine learning models hold promise for clinical decision support, yet their deployment may be limited by the lack of transparent, prediction-level reliability assessment, espec...
BACKGROUND AND OBJECTIVE: Radiomics-based machine learning models hold promise for clinical decision support, yet their deployment may be limited by the lack of transparent, prediction-level reliability assessment, especially under distributional shift. Existing uncertainty estimation methods mainly operate in probability space and may fail to identify unreliable predictions when test samples differ structurally or functionally from the training data. To address this gap, we propose the Radiomics Applicability Domain ANalysis - Classification Framework (RADANCF), a diagnostic approach for assessing the reliability of individual predictions in radiomics classification. METHODS: RADANCF integrates six binary reliability criteria spanning two domains: data representativeness (A-C), describing the relationship between test samples and the training data manifold, and model behavior (D-F), capturing local inconsistencies in predictive responses. Criteria violations are aggregated into ordered reliability categories summarized using a qualitative traffic-light scheme. The framework was evaluated on six public radiomics datasets using five machine learning classifiers, resulting in 900 model configurations trained under a dissimilarity-based stratified partitioning strategy designed to challenge model generalization. Analyses included prediction-level error modeling, multiway ANOVA, correlation analysis between criteria, and assessment of frequently violated criterion combinations. External validation was performed on an independent cohort of 2689 prostate cancer patients from the ProCAncer-I project. RESULTS: Prediction error was significantly associated with RADANCF category, although the relationship was not strictly monotonic, with intermediate categories showing the largest error contributions. RADANCF criteria were largely complementary, as shown by low pairwise Spearman correlations (only 7.5% of cases with correlations higher than 0.5; p < 0.001). Multiway ANOVA confirmed RADANCF category as a significant factor after controlling for dataset and model effects (p < 10⁻¹²). Specific combinations of broken criteria-particularly A, B, C, and E-were significantly overrepresented among higher-error predictions (Wilcoxon test, p < 0.001). In external validation, correct predictions appeared across all traffic-light categories, confirming the diagnostic and risk-oriented nature of RADANCF. CONCLUSIONS: RADANCF provides a transparent, per-prediction diagnostic framework for assessing reliability in radiomics classification under distributional shift. By jointly accounting for data representativeness and model behavior, it complements traditional performance and uncertainty metrics and supports more cautious model deployment in radiomics-based models.
Decoupling lactylation-associated metabolic remodelling in Asthma: Integrated multi-omics and machine learning identify RCC2 as a potential therapeutic target.
Feng Y, Zhou S, Xu J … +1 more , Luo L
Comput Methods Programs Biomed · 2026 Jul · PMID 42000636 · Publisher ↗
BACKGROUND: Aerobic glycolysis promotes lactate accumulation and histone lactylation (Kla), yet the cell-type-specific distribution of this axis and its link to metabolically stratified intercellular signalling in asthma...
BACKGROUND: Aerobic glycolysis promotes lactate accumulation and histone lactylation (Kla), yet the cell-type-specific distribution of this axis and its link to metabolically stratified intercellular signalling in asthma remain poorly defined. We aimed to map lactate/Kla-associated metabolic states across airway lineages and develop an interpretable Lactate Metabolic Biomarker Diagnosis (LMBD) framework. METHODS: By integrating bulk transcriptomes (training: GSE63142; validation: GSE43696) and scRNA-seq (GSE193816), we quantified lactate metabolic activity via four gene sets and 114 metabolic pathways. We identified metabolic subtypes using consensus clustering and trained a Gaussian-mixture-model-guided logistic regression classifier with external validation. Mechanistic insights were inferred through CellChat and Monocle2, followed by functional validation of RCC2 in house dust mite (HDM)-stimulated 16HBE cells using siRNA knockdown, lactate quantification, Pan-Kla immunofluorescence, and untargeted LC-MS metabolomics. RESULTS: Asthma exhibited a macrophage-predominant high-lactate state. We identified a disease-associated macrophage module (Epi1) and two metabolic subtypes, in which the high-lactate group (MBC2) displayed stronger inflammatory transcriptional programs. The LMBD classifier consistently outperformed single-pathway models in the external validation cohort (AUC = 0.799), with RCC2 emerging as the top-ranked feature. Trajectory analysis linked RCC2 to terminal epithelial remodelling. Functional assays demonstrated that RCC2 knockdown significantly reduced intracellular lactate levels, suppressed the expression of glycolysis-associated proteins, and attenuated Pan-Kla signals. Metabolomics further revealed coordinated shifts in redox balance and glycerophospholipid metabolism following RCC2 depletion. CONCLUSION: This integrative study delineates lactate/Kla-associated metabolic remodelling in asthma. The LMBD framework offers a robust transcriptomic proxy for metabolic and inflammatory patient stratification, and RCC2 emerges as a candidate regulatory hub linking glycolysis to epigenetic remodelling, thus warranting further investigation in vivo and in clinical cohorts.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe