Computer Methods And Programs In Biomedicine[JOURNAL]
Sun
200 papers
MMFVS-Net: A triple-symmetric cross-attention network for multimodal optical image fusion and high-accuracy virtual staining of breast cancer tissues.
Shang LW, Guo YX, Li YH … +7 more , Zhu JB, Ji XY, Xiong S, Wu JJ, Huang X, Wang HJ, Yin JH
Comput Methods Programs Biomed · 2026 Jul · PMID 42401051 · Publisher ↗
BACKGROUND AND OBJECTIVE: Virtual staining is based on label-free imaging technology and deep learning, which enables the efficient and non-destructive generation of virtual stained images with effects similar to those o...
BACKGROUND AND OBJECTIVE: Virtual staining is based on label-free imaging technology and deep learning, which enables the efficient and non-destructive generation of virtual stained images with effects similar to those of traditional pathological staining, thus holding great application potential in future clinical medical practice. However, the information provided by a single imaging technology is limited, which is insufficient to support high-accuracy virtual staining. METHODS: This study took breast cancer tissue as the research object. Bright-field, auto-fluorescence and phase contrast imaging technology were introduced to simultaneously extract the multimodal complementary information. A multi-branch input network was established to fuse the three types of optical images and generate high-accuracy virtual stained images. To enhance the model's performance, three strategies were employed, including designing a Triple-symmetric cross-attention module to handle the interactive information of multimodal image features, introducing multi-scale discriminator to ensure the model's ability to capture both global features and local details, and applying transfer-learning to address the gradient dilution issue of multi-branches. RESULTS: Through the evaluation of various image similarity metrics and nuclear statistical metrics, the virtual stained images based on multimodal fusion achieved high accuracy and were significantly superior to any single-modal virtual staining approach. Further ablation experiments explained the reason of the aforementioned results and quantitatively demonstrated the improvements brought by the three optimization strategies. CONCLUSIONS: This work advances the application of multimodal fusion in virtual staining, address the key challenges of cross-modal interactive information mining and multi-branch model training, and further enhance the accuracy and reliability of virtual staining.
A novel Milstein-stochastic epidemiologically-informed neural network for approaching epidemic dynamics: Application to Mpox disease.
Khouna A, Hssayni EH, Joudar NE
Comput Methods Programs Biomed · 2026 Jul · PMID 42398350 · Publisher ↗
BACKGROUND AND OBJECTIVE: Epidemiological dynamics require precise mathematical modeling to guide public health actions, especially for viral diseases such as monkeypox, where data uncertainty and nonlinear transmission...
BACKGROUND AND OBJECTIVE: Epidemiological dynamics require precise mathematical modeling to guide public health actions, especially for viral diseases such as monkeypox, where data uncertainty and nonlinear transmission patterns present significant challenges. In this context, we suggest a novel approach using stochastic epidemic models and deep neural networks. METHODS: In fact, we introduce the Epidemiologically Informed Neural Network (EINN), which uses the classical SIRD model to capture the dynamics of human-to-human transmission of Mpox. Then, we extend to a novel Milstein stochastic epidemiologically informed neural network (M-SEINN), which integrates stochastic differential equations. RESULTS: Our results show that M-SEINN outperformed the Euler SEINN and EINN. At 5% noise in the out-of-sample evaluation, it achieves the lowest RMSE of 3.9569 and the best MAPE of 14.92% for cumulative cases, while at 10% noise in the sample evaluation, the daily case RMSE is 11.29, compared to 12.98 and 14.25, respectively. Statistical analysis demonstrated narrow Bootstrap CIs and a medium-large Cohen's d (0.65-1.02). CONCLUSION: These findings emphasize the necessity of M-SEINN adoption for parameter estimation and public health decisions for epidemic control.
Accounting for approximation errors using surrogate-based parameter estimation of cardiac mechanics digital twins.
Argus F, Creamer SA, Nicholson R … +7 more , Maclaren OJ, Talou GDM, Wang VY, Zhao D, Young AA, Nash MP, Gamage TPB
Comput Methods Programs Biomed · 2026 Jul · PMID 42391886 · Publisher ↗
BACKGROUND AND OBJECTIVE: Parameter estimation for complex physics-based cardiac models is computationally demanding. Surrogate models can be used to speed up model evaluations and improve the feasibility of estimation a...
BACKGROUND AND OBJECTIVE: Parameter estimation for complex physics-based cardiac models is computationally demanding. Surrogate models can be used to speed up model evaluations and improve the feasibility of estimation and uncertainty quantification. However, the use of surrogates introduces additional sources of error that, if neglected, can cause bias or overconfidence in inferences. Here, we present a general approach to account for such model errors when carrying out surrogate-based parameter estimation and uncertainty quantification. METHODS: We use the Bayesian approximation error approach to develop a general framework that systematically accounts for modelling errors and uncertainties induced from the use of a surrogate model. We detail and implement this approach for the task of estimating cardiac stiffness from in-silico 3D left ventricle passive deformation data. We use a finite element model of cardiac mechanics with a neural network-based surrogate, and compare the results with those obtained from a simple regression approach. RESULTS: We show that, despite the sophistication of the neural network, neglecting model errors in the estimation stage leads to biased and overconfident estimates. We demonstrate that our proposed framework allows for simple model-error corrections that provide substantially better inferences. We also demonstrate that our approach can decrease the required number of forward simulations and computational cost for training a neural network by augmenting a low-complexity neural network with a Bayesian approximation error model. CONCLUSIONS: We have developed a framework for augmenting surrogate models that improves inference and decreases training time. This has potential for use in the clinical estimation of cardiac stiffness as a biomarker of disease, where efficiency is required at the point of care.
Facial iPPG heatmap patterns based on period-aware autoencoder show association with carotid atherosclerosis towards non-contact hemodynamic assessment.
Lin Z, Wang Z, Wang H … +3 more , Fang Z, Shen R, Luo J
Comput Methods Programs Biomed · 2026 Jun · PMID 42365708 · Publisher ↗
BACKGROUND AND OBJECTIVE: Early identification of carotid atherosclerosis (CAS) is critical for preventing cardio-cerebrovascular diseases. Mainstream screening methods (e.g. ultrasound, CTA) are operator-dependent and h...
BACKGROUND AND OBJECTIVE: Early identification of carotid atherosclerosis (CAS) is critical for preventing cardio-cerebrovascular diseases. Mainstream screening methods (e.g. ultrasound, CTA) are operator-dependent and high cost. This study aimed to propose a novel non-contact facial imaging photoplethysmography (iPPG) approach for CAS risk assessment. METHODS: A total of 95 middle-aged and elderly participants were enrolled, with synchronous facial iPPG signals and carotid/lower-extremity ultrasound data collected. A deep learning-based Period-aware Autoencoder (PA-AE) with bidirectional cross-modal attention was developed to reconstruct high-fidelity iPPG signals with periodic peak constraint and full-face reference signal fusion for robust noise suppression. Facial hemodynamic heatmaps were generated via signal-to-spatial mapping, interquartile range-based outlier removal, and spatial proximity repair. We analyzed the association between heatmap patterns and atherosclerosis using Pearson chi-square tests and Odds Ratios (OR). RESULTS: The PA-AE outperformed traditional wavelet and LSTM-AE methods in signal periodicity preservation and noise reduction. The Type 3 facial iPPG heatmap (characterized by ≤ 20% red area distributed in the facial periphery) was significantly associated with carotid atherosclerosis (P=0.048), whereas no association was observed for lower extremity atherosclerosis (P=0.674). After adjusting for age, BMI, and hypertension in multivariable logistic regression, heatmap Type 3 still showed a positive trend with CAS (adjusted OR=2.29, 95%CI: 0.56-9.41), and robust statistical analyses including stratified analyses (age < 65 and non-hypertensive subgroups), ridge regression, and continuous red-area ratio quantification consistently confirmed this significant association. CONCLUSIONS: Facial iPPG heatmaps, enhanced by the PA-AE, demonstrate significant potential as a non-invasive tool for identifying CAS risk, offering a promising avenue for accessible community healthcare screening.
Explainable machine learning models predict liver fibrosis risk and outcome in the general population: Development and multi-cohort external validation.
Zhu G, Yi Q, Xu R … +5 more , Xie Y, Song Y, Wang X, Huang L, Xiang Y
Comput Methods Programs Biomed · 2026 Jun · PMID 42365707 · Publisher ↗
BACKGROUND: Significant liver fibrosis is often clinically silent but predicts adverse outcomes. We developed and externally validated an interpretable machine-learning (ML) framework using readily obtainable demographic...
BACKGROUND: Significant liver fibrosis is often clinically silent but predicts adverse outcomes. We developed and externally validated an interpretable machine-learning (ML) framework using readily obtainable demographic, anthropometric, and clinical variables for population-level pre-screening of significant liver fibrosis. METHODS: We included 9424 European participants from the UK Biobank, of whom 1678 were classified as high risk using the Fibrosis-4 Index (FIB-4) ≥ 1.45 or NAFLD Fibrosis Score (NFS) ≥ -1.455. Ten ML algorithms were trained and internally evaluated using a stratified training, validation, and test design. XGBoost was further validated in the 2017-2023 National Health and Nutrition Examination Survey (NHANES; n = 15,270) and an independent real-world cohort (n = 694), in which significant fibrosis was defined as liver stiffness measurement ≥ 8.0 kPa. Discrimination was assessed by the area under the receiver operating characteristic curve (AUC), interpretability by Shapley additive explanations and local interpretable model-agnostic explanations, and prognostic relevance by all-cause mortality. RESULTS: XGBoost achieved AUCs of 0.818 (95% CI, 0.796-0.841) and 0.816 (95% CI, 0.786-0.846) in the internal validation and test cohorts, respectively. In external validation, AUCs were 0.746 (95% CI, 0.735-0.757) in NHANES and 0.793 (95% CI, 0.750-0.836) in the real-world cohort. Interpretation analyses highlighted weight, age, height, hypertension, and waist circumference, with consistent support for anthropometric predictors. Model-defined high-risk status was associated with higher all-cause mortality (log-rank P < 0.001) and remained independently associated with mortality in the internal validation and test cohorts. CONCLUSION: This XGBoost model, based on readily accessible demographic and clinical variables, shows good potential for accurately identifying individuals at high risk of significant liver fibrosis in primary care settings. This approach serves as a valuable tool for improving early detection strategies and facilitating timely interventions for significant liver fibrosis.
Evaluation of surrogate endpoints for survival outcomes using the surrogate package in R.
De Witte D, Molenberghs G, Neyens T … +2 more , Verbeke G, Abad AA
Comput Methods Programs Biomed · 2026 Jun · PMID 42364629 · Publisher ↗
BACKGROUND AND OBJECTIVE: In clinical trials, surrogate endpoints are often used instead of true endpoints due to practical convenience, time efficiency, and reduced sample size requirements. To determine if a surrogate...
BACKGROUND AND OBJECTIVE: In clinical trials, surrogate endpoints are often used instead of true endpoints due to practical convenience, time efficiency, and reduced sample size requirements. To determine if a surrogate endpoint can effectively replace a true endpoint, a meta-analytic approach is often employed, evaluating the surrogate endpoint at both the individual and trial levels. Ideally, both surrogacy levels should be sufficiently high to consider the surrogate endpoint a valid replacement for purposes such as accelerated regulatory approval. We aimed to provide an R implementation for surrogacy evaluation of categorical, survival, or continuous surrogate endpoints for a survival true endpoint. METHODS: In settings where the true endpoint is a time-to-event variable and the surrogate endpoint is continuous, categorical, or also a time-to-event endpoint, the meta-analytic approach to evaluate surrogacy is based on a two-stage approach. In the first stage, a copula model is employed and a measure for individual level surrogacy is estimated. In the second stage, the estimates of the treatment effects are used to compute a measure for trial-level surrogacy. At this step, measurement error can be taken into account via weights or via a model. RESULTS: The R package Surrogate provides functions that implement the two-stage approach for various settings. For all settings, models incorporating Clayton, Plackett and Hougaard copulas are available. Additionally, the functions allow for adjustments for measurement error. We demonstrate their use across these settings by applying them to various datasets. CONCLUSION: We demonstrate how the R package Surrogate can be used to evaluate categorical, survival, or continuous surrogate endpoints for a survival true endpoint.
Relative spectral and frication-based descriptors as numerical indicators of place of articulation shifts in fricatives produced by Polish children.
Skórzewska O, Miodońska Z
Comput Methods Programs Biomed · 2026 Jun · PMID 42364628 · Publisher ↗
BACKGROUND AND OBJECTIVE: The paper undertakes the topic of acoustic differences between normative and pathological articulation patterns in children. The analysis aimed to explore the impact of the place of articulation...
BACKGROUND AND OBJECTIVE: The paper undertakes the topic of acoustic differences between normative and pathological articulation patterns in children. The analysis aimed to explore the impact of the place of articulation on the values of speech signal features in alveolar fricatives / , /. METHODS: Normative (alveolar) pronunciation was compared to three atypical patterns: postalveolar, dental, and interdental. The presented analysis was conducted on our team's dataset of annotated productions collected from 187 preschool-aged children, the largest sample reported to date for the acoustic analysis of pathological child speech. The statistical significance of the differences was assessed using linear mixed-effects models, which account for inter-speaker dependencies in the acoustic data. RESULTS: Our results show that newly introduced relational spectral descriptors, together with noise-energy subbands, reliably capture subtle distinctions among articulatory patterns, whereas fricative formant-level ratios effectively isolate tongue-position effects from voicing. CONCLUSIONS: The study contributes to the field of computer-assisted speech diagnosis by identifying acoustic features to inform future tools dedicated for sigmatism analysis in children.
Leaflet resection improves valve expansion and hemodynamic performance in redo TAVI with balloon- and self-expanding transcatheter heart valve configurations.
Scuoppo R, Cannata S, Gandolfo C … +3 more , Pasquino S, Pasta S, Ferraro M
Comput Methods Programs Biomed · 2026 Jun · PMID 42364627 · Publisher ↗
BACKGROUND AND OBJECTIVE: Redo transcatheter aortic valve implantation (redo-TAVI) is increasingly carried out to treat degenerated transcatheter heart valves (THV). Residual calcified leaflet tissue from the first valve...
BACKGROUND AND OBJECTIVE: Redo transcatheter aortic valve implantation (redo-TAVI) is increasingly carried out to treat degenerated transcatheter heart valves (THV). Residual calcified leaflet tissue from the first valve may impair second-valve expansion and thus prosthesis-patient mismatch and hemodynamic alterations. While leaflet laceration techniques mitigate coronary obstruction, the biomechanical impact of complete leaflet resection remains poorly understood. This study aims to evaluate the structural and hemodynamic effects of transcatheter leaflet resection in redo TAVI for different valve sizes and devices. METHODS: Finite element and smoothed-particle hydrodynamics (SPH) simulations were performed on a patient-specific TAVI model. Redo-TAVI was simulated with and without leaflet resection using both the balloon-expandable Sapien 3 and self-expanding Evolut PRO devices under different THV-in-THV configurations. Expansion and eccentricity indices, effective orifice area (EOA), and mean pressure gradient (PG) were quantified. RESULTS: Leaflet resection consistently improved second-valve expansion and reduced eccentricity, particularly at the annular level. In size-matched balloon-expandable redo-TAVI, annular expansion increased by approximately 30% and eccentricity decreased by nearly 50%. In mismatched size configurations, eccentricity was reduced by >80%. Hemodynamic performance improved accordingly, with EOA increasing by 11-24% and PG decreasing by 21-38% across redo TAVI scenarios. CONCLUSIONS: Findings demonstrated that leaflet resection improves redo-TAVI valve structural and flow performance by reducing asymmetric mechanical constraint from residual leaflet tissue of the degenerated device. The present computational framework provides a valuable tool to guide the optimization of transcatheter system designed to completely resect and remove degenerated valve leaflets prior to redo TAVI.
Spectral super-resolution for Parkinson's voice via representation-level methods under mixed-reality acquisition.
Dudek M, Sikora J, Krzywdziak J … +6 more , Zbik M, Szecowka W, Wodzinski M, Daniol M, Wojcik-Pedziwiatr M, Hemmerling D
Comput Methods Programs Biomed · 2026 Jun · PMID 42361702 · Publisher ↗
BACKGROUND AND OBJECTIVES: Voice is a practical remote biomarker for Parkinson's disease (PD), but real-world capture often yields low-resolution time-frequency inputs that under-resolve diagnostically salient microstruc...
BACKGROUND AND OBJECTIVES: Voice is a practical remote biomarker for Parkinson's disease (PD), but real-world capture often yields low-resolution time-frequency inputs that under-resolve diagnostically salient microstructure. In this controlled empirical comparison, we test whether spectrogram super-resolution (SR) at the feature level performed inside the model rather than via waveform resynthesis improves PD vs. healthy control (HC) discrimination under realistic constraints. METHODS: Speech was recorded with a Microsoft HoloLens 2 using a standardized mixed-reality (MR) protocol from 161 speakers (75 PD/86 HC) across five tasks: Task 1 image description, Task 2 question answering, Task 3 story repetition, Task 4 sustained vowels, and Task 5 word repetition (DDK). Raw audio was exported as 48 kHz, 16-bit PCM, downmixed to mono, amplitude-normalized, conservatively trimmed for leading/trailing silence, and resampled to 16 kHz. Log-mel spectrograms (80 bins) were fed to practical ImageNet-pretrained backbones (ConvNeXt-Tiny, ResNet-50, EfficientNetV2-S). We compared six super-resolution (SR) strategies: identity, nearest (deterministic), bilinear SR, kernel/CARAFE-like SR, LIIF-like SR, and a frozen universal feature SR module (AnyUp) that upsamples intermediate feature maps. Evaluation used 5-fold, speaker-disjoint cross-validation with AUROC (AUC) and accuracy (ACC). RESULTS: AnyUp was the most consistent top performer, ranking first in 11/15 backbone-task cells. Gains were largest on tasks dominated by fine spectro-temporal cues: for ConvNeXt-Tiny, AnyUp vs. identity improved sustained vowels (Task 4) by ΔAUC 0.030/ΔACC 0.070 and DDK (Task 5) by ΔAUC 0.054/ΔACC 0.045. Macro-averaged over tasks, AnyUp outperformed identity by +0.025 AUC/+0.045 ACC (ConvNeXt-Tiny), +0.015/+0.027 (ResNet-50), and +0.052/+0.048 (EfficientNetV2-S). Representative best-in-class results include AUC/ACC of 0.899/0.886 (Task 4) and 0.927/0.897 (Task 5) for ConvNeXt-Tiny+AnyUp, and 0.940/0.903 (Task 5) for ResNet-50+AnyUp. CONCLUSIONS: Densifying spectrogram representations with a frozen, universal feature super-resolution module yields consistent, compute-efficient improvements in PD voice classification under MR-standardized acquisition, with the largest benefits on sustained vowels and DDK. The contribution is empirical rather than architectural: we compare practical representation-level SR choices rather than introduce a new SR module. Feature-level super-resolution is therefore a pragmatic alternative or complement to bandwidth extension when waveform synthesis is unnecessary. At a macro level, the observed accuracy gains (e.g., +0.045 for ConvNeXt-Tiny) suggest that representation-level SR may be operationally useful in low-resolution clinical-audio settings.
Non-invasive coronary fractional flow reserve prediction using a neural network with hemodynamic and geometric embeddings: A proof-of-concept study.
Yang Y, Li B, Fan L … +11 more , Wen C, Liu N, Sun H, Liang LH, Feng Y, Wang T, Liu J, Zhang L, Liu J, Wang L, Liu Y
Comput Methods Programs Biomed · 2026 Jun · PMID 42361701 · Publisher ↗
BACKGROUND AND OBJECTIVES: Invasive fractional flow reserve (FFR) is the clinical gold standard for assessing coronary artery stenosis, but its application is limited by its invasive nature. While CFD-based FFR offers a...
BACKGROUND AND OBJECTIVES: Invasive fractional flow reserve (FFR) is the clinical gold standard for assessing coronary artery stenosis, but its application is limited by its invasive nature. While CFD-based FFR offers a non-invasive alternative, its high computational cost restricts real-time use. Deep learning has emerged as a promising solution. However, many existing methods rely primarily on geometric data and neglect personalized physiological boundary conditions. These limitations hinder the efficient characterization of complex coronary hemodynamics. To address these challenges, this study proposes a method for the rapid, patient-specific prediction of coronary hemodynamics by integrating personalized boundary conditions with geometric information. METHODS: In this study, we constructed a patient-specific steady-state coronary flow field dataset (288 patients). A dual-embedding neural network was proposed, which integrates personalized hemodynamic constraints with coronary geometry to enhance predictive accuracy. The network employs dual encoders to separately extract features from personalized hemodynamic boundary conditions and reduced-dimensional coronary geometry. By integrating these encoded representations into a Bi-LSTM architecture, the model learns the relationship between vascular topology and hemodynamic parameters. The predicted FFR values were validated against both CFD-based FFR and invasive FFR to assess numerical agreement and diagnostic performance. RESULTS: On the independent test set, the model achieved an RMSE of 7.42% compared to CFD-based FFR. When validated against invasive FFR, the FFR showed a correlation of 0.87 and an AUC of 0.901. The inference time per case was significantly reduced compared to traditional CFD. CONCLUSIONS: The dual-embedding neural network achieves high consistency with both CFD-based FFR and invasive FFR. By encoding patient-specific coronary structures and boundary conditions, this approach provides a proof of concept that FFR distributions can be computed in real time, even with a relatively small CFD training dataset. These findings demonstrate a robust, high-efficiency solution with significant potential for broader application in diverse cardiovascular scenarios.
Comment on "Facial surgery preview based on the orthognathic treatment prediction" (Han et al., 2025).
Lien KH
Comput Methods Programs Biomed · 2026 Jun · PMID 42361700 · Publisher ↗
Abstract loading — click title to view on PubMed.
AneuSI: A software tool to isolate a region of interest in cerebral aneurysms from full geometrical models.
Muzi N, Moyano LG, Millán D
Comput Methods Programs Biomed · 2026 Jun · PMID 42348982 · Publisher ↗
BACKGROUND AND OBJECTIVE: Modeling cerebral aneurysms using patient-specific geometries demands significant computational resources, particularly when analyzing large datasets or to retrieve training data for machine lea...
BACKGROUND AND OBJECTIVE: Modeling cerebral aneurysms using patient-specific geometries demands significant computational resources, particularly when analyzing large datasets or to retrieve training data for machine learning techniques. Smaller domains representing the aneurysm and nearby vasculature are preferred to reduce computational cost, but manual extraction of these regions introduces limitations in time, repeatability, and user-dependent bias. In this work, we present and study the accuracy of AneuSI, an open-source software tool designed to automatically extract a region of interest around the aneurysm from the arterial tree without requiring user intervention. The tool is specifically optimized for the AneuriskWeb database format and its centerline data structure, due to its extensive case collection and high quality, curated data. While its modular structure enables future adaptation to other datasets, the current implementation and validation focus on this specific format. METHODS: AneuSI is developed in C++ and based on the open-source Visualization Toolkit library. It operates as a command-line tool for Linux, supporting automation through bash scripting. Isolation length is defined as the multiple of a clip factor k and the vessel's inner diameter, obtaining results relative to each aneurysm dimensions. Clip quality was evaluated by comparing AneuSI's output against a trained operator's results for 10 patient-specific lateral aneurysm geometries. The tool was further challenged by processing 102 aneurysm cases from 99 patients across 7 different k values. RESULTS: AneuSI achieved results comparable to manual isolation in a fraction of the time, averaging 2 s per case in a single processor desktop PC, compared to the 10-15 min needed for manual processing. Applied to the AneuriskWeb database, AneuSI demonstrated a 100% success rate, obtaining 714 isolated models through a total of 2592 cuts. CONCLUSION: AneuSI provides a fast, robust, and reproducible solution for automatic aneurysm-centered region extraction, reducing variability while drastically improving efficiency. Its modular architecture enables straightforward integration into existing tools or libraries, and its compatibility with batch processing make it suitable for large-scale biomechanical studies and machine learning dataset generation in cerebrovascular modeling. Certain geometric heuristics were prioritized for computational efficiency; robustness on heterogeneous datasets or alternative centerline formats requires further adaptation.
Deep learning-based segmentation of human oocytes with cross-dataset evaluation.
Lei Z, Xu X, Tilia L … +6 more , Toit-Thompson SD, Horta F, Gilchrist RB, Walls ML, Bhargava A, Goldys EM
Comput Methods Programs Biomed · 2026 Jun · PMID 42341360 · Publisher ↗
BACKGROUND AND OBJECTIVE: Accurate segmentation of human oocyte subregions in microscopy images is an important step towards developing objective and automated systems to support oocyte quality assessment in assisted rep...
BACKGROUND AND OBJECTIVE: Accurate segmentation of human oocyte subregions in microscopy images is an important step towards developing objective and automated systems to support oocyte quality assessment in assisted reproductive technologies. This study aims to segment four clinically relevant oocyte regions-ooplasm, perivitelline space (PVS), zona pellucida (ZP), and first polar body (PBI)-using deep learning models and to evaluate their performance across multiple clinical datasets. METHODS: Four segmentation models (UNet++, DeepLabV3+, SegFormer, and a transformer-based architecture) were trained on a clinically annotated dataset. The models were evaluated using the Intersection Over Union (IoU), Dice Similarity Coefficient (DSC), precision, and recall. To assess performance, external datasets collected under different imaging conditions from other clinical sites were used for testing. Domain adaptation techniques were applied by fine-tuning the pretrained models on external clinical datasets to improve cross-site robustness. RESULTS: SegFormer achieved the highest segmentation performance, consistently outperforming convolutional neural network (CNN) models in oocyte segmentation. On the primary training dataset, the SegFormer model achieved high DSC across all regions, with mean values exceeding 99% for ooplasm, 87% for PVS, 94% for ZP, and 86% for PBI. Fine-tuned models pretrained on a single oocyte dataset with external clinical images yielded better performance than ImageNet-based initialisation, highlighting the advantage of task-specific pretraining. CONCLUSIONS: These findings confirm the feasibility of multi-component oocyte segmentation and establish a basis for future quality assessment studies.
SPELL: A scalable NLP method using regular expressions and large language models for clinical information extraction.
Kleinlein R, Bates DW, Guan C … +2 more , Gray KJ, Kovacheva VP
Comput Methods Programs Biomed · 2026 Jun · PMID 42341359 · Publisher ↗
BACKGROUND AND OBJECTIVES: Electronic health records (EHRs) contain valuable information for research and decision-making, but much resides in unstructured notes that are challenging to analyze at scale. We developed SPE...
BACKGROUND AND OBJECTIVES: Electronic health records (EHRs) contain valuable information for research and decision-making, but much resides in unstructured notes that are challenging to analyze at scale. We developed SPELL (Snippet-Primed rEgex LLM Pipeline), a scalable natural language processing workflow that combines regular-expression-based snippet retrieval with locally hosted large language model (LLM) inference to extract structured variables from large collections of clinical narratives. METHODS: SPELL uses task-specific regular expressions to retrieve short context windows ("snippets") from unstructured texts and applies task-prompted LLM inference on snippets rather than full documents. All processing occurs within institutional computing environments. Accuracy was evaluated on randomly sampled, clinician-annotated benchmark sets of 50 documents per obstetric task, with separate retrieval-recall audits of 20 regex-negative documents per task. We evaluated accuracy and efficiency across three obstetric information-extraction tasks: numerical value (blood loss volume), date (estimated due date), and diagnosis (hemolysis, elevated liver enzymes, and low platelets [HELLP] syndrome). We quantified computational scalability using elapsed time, out-of-memory events, energy consumed, and GPU telemetry, and audited retrieval recall using clinician-annotated regex-negative notes enriched with relevant structured metadata. Generalizability was assessed on the public MT Samples corpus (5013 notes across 40 specialties) for ventricular tachycardia detection. RESULTS: SPELL processed 31 million clinical notes spanning 1976-2024 from eight hospitals. Snippet-based inference reduced processing time by 71-87% versus full-document LLM inference and by >95% versus manual physician annotation. On the 50-document benchmark sets, snippet-based evaluation achieved 98% exact-match accuracy for blood-loss extraction, 92% exact-match accuracy for estimated-due-date extraction, and 94% accuracy with an F1-score of 0.97 for HELLP syndrome classification. As an exploratory external evaluation on MT Samples, ventricular tachycardia detection achieved 84% accuracy and an F1-score of 0.67. CONCLUSIONS: A hybrid regex-snippet-LLM pipeline can enable accurate and computationally efficient extraction from unstructured EHR narratives.
Deep learning based hair removal on ultraviolet-induced fluorescence dermatoscopy images.
Kempski A, Salwowska N, Slian A … +5 more , Korecka K, Pietkiewicz P, Polańska A, Borys D, Czajkowska J
Comput Methods Programs Biomed · 2026 Jun · PMID 42330594 · Publisher ↗
BACKGROUND AND OBJECTIVE: Ultraviolet-induced fluorescence dermatoscopy (UVFD) is a novel technique utilizing ultraviolet light to produce visible fluorescence. Many skin conditions reveal specific features during UVFD,...
BACKGROUND AND OBJECTIVE: Ultraviolet-induced fluorescence dermatoscopy (UVFD) is a novel technique utilizing ultraviolet light to produce visible fluorescence. Many skin conditions reveal specific features during UVFD, guiding early detection and diagnosis. Although there are many methods for computer processing of dermatoscopic images, there are no methods targeting UVFD images. This work identifies newly encountered challenges and proposes a preprocessing pipeline for preparing obtained data for further statistical analysis. METHODS: The study discusses the challenges and the applicability of methods developed for regular dermatoscopic images. Finally, the universal pipeline for preprocessing is presented. The study used a collection of 2027 background images, manually selected by an expert from hairless regions, which were later used for hair generation. Additionally, 120 images were manually annotated for dark and light hair, resulting in a set of ground truth masks. The study evaluated a number of algorithmic methods and a deep neural networks: UNet, ChimeraNet and TransUNet. RESULTS: The results clearly showed the superiority of deep learning over algorithmic methods in the task of hair segmentation in UVFD images. For dark hair segmentation, the UNet and TransUNet reported Dice scores ranging 0.95-0.98 on synthetic data. On real-life light hair UNet achieved the higher result, with Dice score of 0.557, while on dark hair all models ranged 0.61-0.65. CONCLUSIONS: The study used a synthetic dataset to train the UNET neural network, which yielded results superior to analytical approaches known from dermatological images. The presented preprocessing pipeline is freely available and can be found on our GitHub repository https://github.com/aKempski01/UVFD-data-preprocessing.
Data-driven system identification in cancer systems biology: A multiscale modeling approach to melanoma.
Christenson C, Viknesh S, Judson-Torres RL … +1 more , Arzani A
Comput Methods Programs Biomed · 2026 Jun · PMID 42322753 · Publisher ↗
BACKGROUND AND OBJECTIVE: Developing systems biology models of cancer is critical for improving understanding of complex biological interactions and advancing appropriate therapeutic strategies. However, constructing com...
BACKGROUND AND OBJECTIVE: Developing systems biology models of cancer is critical for improving understanding of complex biological interactions and advancing appropriate therapeutic strategies. However, constructing complex, multiscale models that are easily parameterized by the available data can be difficult. Advances in computational methods have provided data-driven frameworks utilizing machine learning approaches to build predictive systems biology models. These methods display a strong ability to fit data and predict cellular and molecular trajectories but can be limited by either poor mechanistic interpretations or require a large prior understanding of the system of interest. In many cases, such as cancer, the systems are highly complex, resulting in limited availability of prior knowledge and a strong need for mechanistic insight from the developed models. This work builds upon recent developments in system identification to construct fully mechanistic models directly from the available data, serving as a proof-of-concept study for multi-scale system identification with synthetic data. METHODS: We utilize ADAM-SINDy (sparse identification of nonlinear dynamics with ADAM) to accomplish this objective, a recently developed differentiable optimization method for the identification of nonlinear, parameterized dynamical systems, with advancements to facilitate its use in systems biology settings. Specifically, we focus on the usage of prior biological knowledge and the subsequent finetuning through a two-stage identification process. We apply the framework in the context of melanoma (a dangerous skin cancer) and quantitative systems pharmacology, developing a benchmark multiscale model connecting sub-cellular signaling and cell scale dynamics in response to therapy. RESULTS: The ADAM-SINDy framework is shown to be capable of identifying a ground truth system of equations directly from synthetically generated data. Coupling mechanisms between scales (subcellular and cellular) are also identifiable with this framework. At high temporal resolutions, the method achieves 100% reconstruction accuracy, dropping to 89.9% with a two times coarser resolution. CONCLUSIONS: The framework presented in this work is capable of constructing systems biology networks and mathematical models directly from data, while maintaining clear mechanistic interpretations. This system identification paradigm in systems biology can help remove the need for extensive biological insight into the network of interest a priori, which can be exceedingly difficult in the context of cancer.
Dual-CNN-based AI framework for patient-specific design of short femoral stems.
Yáñez González-Cuéllar JA, Moscol Albanil IDP, Ojeda C … +3 more , Franco-Martínez F, Díaz Lantada A, Solórzano-Requejo W
Comput Methods Programs Biomed · 2026 Jun · PMID 42320193 · Publisher ↗
This study introduces an AI-driven methodology for the personalized design of short femoral stems in total hip arthroplasty, addressing the challenge of stress shielding that compromises long-term implant survival. To im...
This study introduces an AI-driven methodology for the personalized design of short femoral stems in total hip arthroplasty, addressing the challenge of stress shielding that compromises long-term implant survival. To improve pre-surgical implant suitability, a dual-input convolutional neural network (dual-CNN) was developed to predict the shielding directly from CT-like cross-sectional images that simultaneously capture femoral anatomy and stem geometry. A digitally generated dataset based on two segmented femurs and 392 stem designs was used for training and validation, while a third unseen femur assessed generalization. The influence of different dataset configurations was analyzed, with the combined dataset yielding the most accurate and robust predictions. The dual-CNN outperformed both single-anatomy models and a previously published random forest approach, reducing mean absolute error by approximately 30% and confirming the benefits of anatomically informed, image-based inputs. These findings demonstrate that the proposed model offers an efficient and scalable alternative to finite element analysis for evaluating stress/strain shielding and optimizing patient-specific short femoral stem designs.
DDVMM: A dual-branch pyramid model for mono-modal medical image registration.
Zou M, Wu Y, Chen J … +1 more , Liu Q
Comput Methods Programs Biomed · 2026 Jun · PMID 42320192 · Publisher ↗
BACKGROUND AND OBJECTIVE: Medical image registration is pivotal in clinical diagnosis and image-guided intervention. However, convolutional neural network (CNN)- and Transformer-based approaches still face challenges in...
BACKGROUND AND OBJECTIVE: Medical image registration is pivotal in clinical diagnosis and image-guided intervention. However, convolutional neural network (CNN)- and Transformer-based approaches still face challenges in achieving high registration efficiency and accuracy simultaneously, particularly when dealing with large deformations and significant anatomical differences. Therefore, a more robust and efficient registration framework is highly desirable for medical image analysis. METHODS: We propose DDVMM, a dual-branch pyramidal registration model integrating Transformer and State Space Models (SSMs), to enhance multi-scale feature extraction and global structural perception. The encoder incorporates a Directional Depthwise Convolution (DDC) module and a Deformable Multi-Head Self-Attention (D-MHSA) module to enhance local boundary and directional feature extraction, as well as long-range spatial dependency modeling. The decoder employs a State Space Model-based guided feature interaction module (SSM Decoder) that captures large-scale spatial context through hidden state propagation and fuses multi-scale features to improve structural alignment and detail refinement. The network adopts a context-guided coarse-to-fine deformation estimation strategy, progressively refining deformation predictions to improve overall registration accuracy. RESULTS: Extensive experiments are conducted on two public datasets: the ACDC cardiac dataset and the OASIS-1 brain dataset. The results demonstrate that the proposed method achieves higher average Dice similarity scores and lower average boundary errors compared with the competing methods. CONCLUSIONS: This work presents a novel dual-branch pyramidal registration framework that effectively combines Transformer and State Space Models(SSMs) to enhance both global structural perception and local detail alignment. The proposed DDVMM achieves better registration performance in handling large deformations and anatomical differences while using fewer parameters, resulting in a better performance-parameter trade-off. Code is available at https://github.com/wyl32123/DDVMM.
Clinically-aligned explainable AI for atrial fibrillation detection: A U-Net inspired multi-lead ECG analysis framework.
Taleban A, Sparapani R, Zlochiver S … +3 more , Lu Q, Widlansky ME, Luo J
Comput Methods Programs Biomed · 2026 Jun · PMID 42314562 · Publisher ↗
BACKGROUND AND OBJECTIVE: Deep learning offers high diagnostic accuracy but often lacks interpretability, hindering clinical adoption for Electrocardiogram (ECG) analysis. This study aims to develop a clinically interpre...
BACKGROUND AND OBJECTIVE: Deep learning offers high diagnostic accuracy but often lacks interpretability, hindering clinical adoption for Electrocardiogram (ECG) analysis. This study aims to develop a clinically interpretable, U-Net-inspired deep learning framework for Atrial Fibrillation (AFib) detection that elucidates model decision-making through morphological and rhythmic feature extraction. METHODS: We utilized a U-Net-inspired encoder-decoder architecture to analyze 12-lead ECGs as spatially structured inputs, capturing inter-lead relationships using the PTB-XL database. Model interpretability was assessed using layer-wise Gradient-weighted Class Activation Mapping (Grad-CAM) to visualize hierarchical feature focus. Additionally, we implemented RR interval variability assessment with heart rate stratification and group-specific threshold optimization to account for physiological variations. RESULTS: On PTB-XL-Improved labels, the model achieved an Area Under the Curve (AUC) of 99.15%, Sensitivity of 97.81%, Specificity of 95.63%, and Precision (PPV) of 95.72%. Performance on the original PTB-XL annotations reached an AUC of 99.78%. Grad-CAM analysis confirmed hierarchical feature extraction, correctly localizing QRS complexes in over 97.5% of cases and P-wave and rhythm irregularities in over 93.1% of cases. Independent RR-interval variability analysis achieved AUCs of 95.5% (slow heart rate), 93.2% (normal), and 91.3% (fast) across three clinically stratified heart rate groups. CONCLUSIONS: This framework successfully combines high diagnostic performance with transparent, clinically aligned visual explanations. The principal contribution is not accuracy alone, but the demonstration that layer-wise explanations can be aligned with the clinical ECG reading sequence, from QRS localization to P-wave morphology. External validation remains necessary before broader clinical generalization can be claimed.
A novel 0D-1D-3D multi-scale blood flow model: Application to atherosclerotic plaque hemodynamics.
Wang D, Chen H, Ma Z
Comput Methods Programs Biomed · 2026 Jun · PMID 42314561 · Publisher ↗
OBJECTIVE: Atherosclerosis (AS) poses a significant threat to human health. Its early and efficient diagnosis is still limited by the lack of research tools, which has led to an unclear understanding of pathological bloo...
OBJECTIVE: Atherosclerosis (AS) poses a significant threat to human health. Its early and efficient diagnosis is still limited by the lack of research tools, which has led to an unclear understanding of pathological blood flow characteristics. This study aims to further improve the hemodynamic models for exploring the universal pathological flow features of plaques, and to lay the foundation for a hemodynamics-based strategy for AS identification. METHODS: A novel 0D-1D-3D coupled multi-scale model is proposed. Employing a 0D three-element Windkessel model as the distal boundary condition, it obtains an analytical solution for the major arteries in the frequency-domain. Based on a fluid-structure interaction approach, the detailed 3D local blood flow in key arterial regions is captured in the time-domain. Within the algorithm framework of time-frequency iterative solution, a physiologically consistent coupling between the 1D and 3D models is achieved. Based on medical imaging data, the plaque is modeled parametrically. As a case study, the multi-scale model is used to compare the influence of various plaque geometrical parameters on systemic hemodynamics from a frequency-domain perspective, and to analyze their corresponding flow patterns. RESULTS: Compared to full 3D results for the abdominal aortic bifurcation case, the proposed model improves computational efficiency by 55%, with maximum pressure and flowrate root-mean-square errors of 1.47% and 2.95%, respectively. In the presence of arterial stenosis, monitoring the 5th to 8th harmonic pressure amplitudes and 4th to 6th harmonic phases in the frequency-domain has the potential to identify plaque geometrical features. Blood flow exhibits the strongest sensitivity to the vertical radial radius of the plaque and the weakest sensitivity to its length. CONCLUSION: The proposed model can accurately and effectively simulate blood flow in the arterial system, and offer a novel insight for the identification strategy, regarding the severity and basic geometry of arterial stenosis based on hemodynamic frequency-domain features.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe