J Assist Reprod Genet. 2025 Aug 12. doi: 10.1007/s10815-025-03570-x. Online ahead of print.

ABSTRACT

OBJECTIVE: To evaluate whether the use of a fully automated AI-based scoring system (iDAScore V2) for selecting viable embryos using fetal heartbeat (FHB) as an indicator is equivalent to morphology assessment.

METHODS: A retrospective observational cohort study across four fertility centers analyzed embryos selected for single embryo transfer on Day 3 or Day 5 + based on morphology and time-lapse video. All transferred embryos from participating centers were retrospectively scored using a fully automated AI-based embryo scoring algorithm and standardized morphology assessment. The predictive ability of both methods for implantation (FHB rate) was compared for Day 3 and Day 5 + transfer.

RESULTS: A multi-center analysis revealed that AI-based embryo scoring significantly outperformed morphological embryo assessment in predicting FHB for both Day 3 (n = 2965) and Day 5 + (n = 6970) transfers (P < 0.0001). Similarly, the discrimination of low versus high scores regarding FHB resulted in a significantly better area under the curve (AUC) for iDAScore V2 compared to standardized morphology assessment for Day 3 (0.63; 95% CI: 0.61-0.65 versus 0.59; 95% CI: 0.58-0.61) and for Day 5 + (0.59; 95% CI: 0.57-0.60 versus 0.55; 95% CI: 0.54-0.57).

CONCLUSIONS: As a multi-center validation of fully automated embryo assessment, this study confirms that AI-based selection provides outcomes that are either equivalent to or superior to morphological embryo assessment, without compromising clinical outcomes.

PMID:40794157 | DOI:10.1007/s10815-025-03570-x

Neuroradiology. 2025 Aug 12. doi: 10.1007/s00234-025-03733-8. Online ahead of print.

ABSTRACT

PURPOSE: This study compares a novel reconstruction algorithm deep learning-based image reconstruction (DLIR) and adaptive statistical iterative reconstruction-Veo (ASIR-V) for CTA in acute ischemic stroke (AIS) patients, emphasizing DLIR's potential to improve diagnostic accuracy and visualization of large vessel occlusion.

METHODS: This study retrospectively assessed 108 consecutive AIS-suspected emergency department patients (mean age 72.3 years +/- 17) who underwent head and neck CTA with DLIR and ASIR-V reconstructions. The analysis compared the impact of DLIR versus ASIR-V on image quality, assessing signal-to-noise (SNR), contrast-to-noise ratios (CNR), and contrast-enhanced arteries homogeneity computed on mean HU values and SD in six regions of interest located in head and neck including three arteries.

RESULTS: The DLIR reconstruction allowed for significant SNR and CNR improvement, with the largest SNR distinction obtained in the common carotid artery (52.29% increased SNR) and white matter of the pons (63.98% increased SNR). Among the three regions subject to CNR evaluation DLIR yielded superiority in the neck and posterior cerebral fossa while ASIR-V accounted for higher CNR in the medial cerebral fossa (MCF). Additionally, DLIR-reconstructed images achieved a 21.10% improvement in arterial homogeneity, enhancing the visualization of potential occlusion.

CONCLUSION: DLIR yields superior image quality of the contrast-enhanced head and neck structures in CTA, providing artery images with increased homogeneity and potentially allowing for more proficient occlusion evaluation specifically in the area of the posterior cerebral fossa. However, this technique faces challenges in the visualization of MCF.

PMID:40794135 | DOI:10.1007/s00234-025-03733-8

Biopreserv Biobank. 2025 Aug 12. doi: 10.1177/19475535251367317. Online ahead of print.

ABSTRACT

Introduction: This study is part of the broader Stem Line project Mito-Cell-UAB073, specifically focusing on "Stem Cell Lines-Quality Control," and aims to innovate in the field of Quality Control (QC) through a unique, artificial intelligence (AI)-powered model known as Life Cell AI UAB. This model utilizes deep learning algorithms and computer vision, allowing it to make accurate viability assessments of cell and stem cell lines based solely on static images captured through standard optical microscopes. Aim: The aim of this study was to develop and validate an AI-driven, image-based model that reliably predicts cell line viability. Methods: Our methodology involved training the Life Cell AI UAB model on single static images of cell lines using advanced computer vision and deep learning techniques. Performance evaluation was conducted on three independent blind test sets sourced from various biotechnology laboratories, allowing for assessment across diverse environments. Results: The Life Cell AI UAB model achieved a sensitivity of 82.1% in identifying viable cell lines and a specificity of 67.5% for non-viable lines across the test sets. Each blind test set exhibited a weighted accuracy above 63%, with a combined accuracy of 64.3%. Notably, predictions showed a clear distinction between correctly and incorrectly classified cells. The model outperformed traditional QC methods by improving accuracy in binary classification tasks by 21.9% (p = 0.042) and demonstrated a 42.0% enhancement over conventional Standard Operation Procedure (SOP) procedures (p = 0.026). Conclusion: The Life Cell AI UAB model represents a notable advancement in biobanking QC, offering a precise, standardized, and non-invasive method for assessing cell line viability. This model has the potential to streamline QC processes across laboratories, minimizing the need for time-lapse imaging and promoting uniformity in QC practices for both cell and stem cells.

PMID:40793964 | DOI:10.1177/19475535251367317

Int J Surg. 2025 Aug 11. doi: 10.1097/JS9.0000000000003182. Online ahead of print.

NO ABSTRACT

PMID:40793836 | DOI:10.1097/JS9.0000000000003182

Metabolomics. 2025 Aug 12;21(5):111. doi: 10.1007/s11306-025-02316-1.

ABSTRACT

INTRODUCTION: Myocardial revascularization (MR) is recommended in acute myocardial infarction. Understanding the physiological disturbances caused by MR may be pertinent for future therapeutic strategies in the postoperative period.

OBJECTIVES: This study aims to analyze the MR impact on plasma metabolites and investigate potential correlations between them with routinely measured biochemical and clinical parameters in MR, and with the cardiopulmonary bypass time (CPB).

METHODS: Twenty-five patients had plasma samples collected before and after MR for metabolomic analysis, performed by liquid chromatography coupled with high-resolution mass spectrometry.

RESULTS: One hundred eleven ions showed statistically significant differences due to MR and thirteen were identified. Only Pregnenolone Sulfate had its abundance in plasma decreased due to MR. Hydrocortisone succinate, Cortisone, and Corticosterone increased in response to the glucocorticoids administered during surgery. LysoPS 16:1, LysoPC 14:0, Phenylvaleric acid, 13-Hydroxyoctadecadienoic acid, N-Linoleoyl Glutamine, and N-Myristoyl Methionine, along with the significant increase in the white blood cell count are associated with inflammation processes, possibly caused by MR. Furthermore, Pregnenolone sulfate, Pentosidine, Phenylvaleric acid, and N-Myristoyl Methionine were correlated with biochemical/clinical parameters and CPB.

CONCLUSION: These results open new horizons in the interpretation of physiological disturbances caused by MR, as well as provide support for future studies. The scientific community is invited to build upon the outcomes obtained to confirm the associations suggested in this study, advancing the realm of metabolomics and MR.

PMID:40794378 | DOI:10.1007/s11306-025-02316-1

Metabolomics. 2025 Aug 12;21(5):112. doi: 10.1007/s11306-025-02310-7.

ABSTRACT

INTRODUCTION: Rheumatoid arthritis (RA) is an auto-immune disease which causes irreversible damage to tissue and cartilage within synovial joints. Rapid diagnosis and treatment with disease-modifying therapies is essential to reduce inflammation and prevent joint destruction. RA is a heterogeneous disease, and many patients do not respond to front-line therapies, requiring escalation of treatment onto biologics, of which TNF inhibitors (TNF-i) are the most common.

OBJECTIVES/METHODS: In this study we determined whether serum metabolomics, using nuclear magnetic resonance (NMR) and Fourier transform infrared (FTIR) spectroscopy, could discriminate RA blood sera from healthy human controls and whether the technologies could be used to predict response or non-response to TNF inhibitor (TNF-i) therapy.

RESULTS: NMR spectroscopy identified 35 metabolites in RA sera, with acetic acid being significantly lower in RA sera compared to healthy controls (HC, FDR < 0.05). PLS-DA modelling identified 2-hydroxyisovalericacetic acid, acetoacetic acid, mobile lipids, alanine and leucine as important metabolites for discrimination of RA and HC sera by 1H NMR spectroscopy (averaged 83.1% balanced accuracy, VIP score > 1). FTIR spectroscopy identified a significant difference between RA and HC sera in the 1000-1200 cm- 1 spectral area, representing the mixed region of carbohydrates and nucleic acids (FDR < 0.05). Sera from RA patients who responded to TNF-i were significantly different from TNF-i non-responder sera in the 1600-1700 cm- 1 region (FDR < 0.05).

CONCLUSION: We propose that NMR and FTIR serum metabolomics could be used as a diagnostic tool alongside current clinical parameters to diagnose RA and to predict whether someone with severe RA will respond to TNF-i.

PMID:40794325 | DOI:10.1007/s11306-025-02310-7

Transl Cancer Res. 2025 Jul 30;14(7):4179-4193. doi: 10.21037/tcr-2024-2563. Epub 2025 Jul 23.

ABSTRACT

BACKGROUND: Neuroblastoma is a highly heterogeneous pediatric malignancy, with high-risk cases exhibiting poor clinical outcomes. Store-operated calcium entry (SOCE) channels have been implicated in cancer progression, yet their prognostic significance in neuroblastoma remains unclear. This study aimed to investigate the relevance of SOCE-related genes in predicting patient prognosis and guiding therapeutic strategies.

METHODS: We performed unsupervised clustering based on SOCE-related gene expression in multiple neuroblastoma RNA sequencing (RNA-seq) datasets. A prognostic scoring system, the SOCE_Score, was developed using machine learning algorithms. The model's predictive performance was validated across independent datasets. Immune characteristics were assessed using established deconvolution algorithms, and candidate therapeutic compounds were identified via the Connectivity Map (CMap) platform.

RESULTS: Two distinct molecular clusters were identified, differing significantly in survival outcomes, immune infiltration, and stemness signatures. The SOCE_Score stratified patients with high accuracy and outperformed conventional clinical predictors. Lower SOCE_Score groups were associated with favorable immune landscapes and greater responsiveness to immune checkpoint blockade. CMap analysis highlighted MS-275, a histone deacetylase (HDAC) inhibitor, as a promising compound targeting low SOCE_Score phenotypes.

CONCLUSIONS: SOCE-related transcriptional features serve as robust biomarkers for prognosis and immune activity in neuroblastoma. The SOCE_Score holds potential for guiding risk stratification, immunotherapeutic selection, and drug repurposing efforts. These findings underscore the clinical utility of integrating calcium signaling profiles into neuroblastoma management and warrant further experimental validation.

PMID:40792155 | PMC:PMC12335702 | DOI:10.21037/tcr-2024-2563

Biotechnol Rep (Amst). 2025 Jul 29;47:e00909. doi: 10.1016/j.btre.2025.e00909. eCollection 2025 Sep.

ABSTRACT

Alzheimer's disease affects 6.9 million Americans aged 65 and older, a number expected to double by 2060. Eight FDA-approved drugs target Alzheimer's, but no cure is available, and most treatments are symptomatic. Drug repurposing, the use of FDA-approved drugs for new indications, is a promising strategy to address this lack of effective therapies. However, despite prior safety approval, repurposable drugs may still trigger unexpected side-effects in new contexts. This study introduces a network-based approach to minimize side-effect risk in drug repositioning, focusing on QT interval prolongation, a cardiac side-effect observed in Alzheimer's patients treated with acetylcholinesterase inhibitors. The method integrates Mode-of-Action and Random Walk with Restart analyses to identify repositioning candidates while assessing QT-related risk. This strategy identified promising compounds including acamprosate, tolcapone, sitagliptin, and diazoxide, with potential to mitigate disease pathology. Gene set enrichment analysis was used to computationally assess the compounds' ability to reverse disease-related gene expression signatures.

PMID:40791766 | PMC:PMC12337645 | DOI:10.1016/j.btre.2025.e00909

BMC Med Inform Decis Mak. 2025 Aug 11;25(1):298. doi: 10.1186/s12911-025-03146-w.

ABSTRACT

BACKGROUND: Data quality (DQ) of electronic health record (EHR) is crucial for the advancement of health informatization, yet it remains a significant challenge. Scholars are showing a growing interest in leveraging semantic technologies to enhance EHR data quality. However, previous studies have focused predominantly on specific semantic technologies, scenarios, or objectives-such as interoperability-often overlooking the potential of a various semantic technologies across different scenarios.

OBJECTIVE: This systematic review aimed to explore the potential of employing a range of semantic technologies to improve EHR data quality in a broader spectrum of application scenarios.

METHODS: Our systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Three databases were searched, including PubMed, IEEE Xplore, and Web of Science Core Collection. The search terms used included "Semantic*", "Quality", "Electronic Health Record*", "EHR*", "Electronic Medical Record*", and "EMR*". These terms were combined via various Boolean operators to formulate multiple search queries.

RESULTS: Thirty-seven papers that met the inclusion criteria between 2008 and 2024 were analyzed. Six semantic techniques were identified as instrumental in improving EHR DQ: EHR standardization, controlled vocabulary, ontology, semantic web, knowledge graph, and natural language processing (NLP). These technologies were further mapped to 16 core data quality indicators and the FAIR principles (Findable, Accessible, Interoperable, and Reusable), highlighting their contributions across both technical and governance dimensions.

CONCLUSIONS: The six identified semantic technologies can be categorized into three levels: foundational, general, and advanced. These technologies show significant potential in enhancing EHR DQ, particularly in the areas of conformance, portability, usability, and applicability, and they are suitable for a variety of contexs beyond interoperability, aligning with FAIR-aligned best practices in data management and reuse.

PMID:40790196 | DOI:10.1186/s12911-025-03146-w

Front Cardiovasc Med. 2025 Jul 28;12:1546493. doi: 10.3389/fcvm.2025.1546493. eCollection 2025.

ABSTRACT

BACKGROUND: Phenotypic heterogeneity is apparent among individuals with putative monogenic disease, such as familial hypertrophic cardiomyopathy. Genome sequencing (GS) allows interrogation of the full spectrum of inborn genetic variation in an individual and RNA profiling provides a snapshot of the cardiac-specific pathogenic effects on gene expression.

OBJECTIVES: Identify candidate genetic modifiers of hypertrophic cardiomyopathy phenotype.

METHODS: We performed GS of 48 individuals with variants in MYH7, the gene encoding beta myosin heavy chain, and a personal or family history of cardiomyopathy. The genome sequences were annotated with a custom pipeline optimized for cardiovascular gene variant detection. We utilized multiple lines of evidence to prioritize genes together with rare variant gene-based association testing to identify candidate genetic modifiers.

RESULTS: GS identified the MYH7 variant in all 48 cases. Several variants were reclassified based on best available data. We identified known disease-associated genes (MYBPC3, FHOD3), a priori candidate modifiers (ATP1A2, RYR2), and novel candidate modifiers of cardiomyopathy including PACSIN3 and SORBS2. We identified regulatory variants and intergenic regions associated with the phenotypes. Using RNA profiling, we show that several genes identified through gene-based association testing are differentially regulated in human hypertrophic cardiomyopathy, and in models of disease.

CONCLUSION: Evaluation of the whole genome, even in the case of alleged monogenic disease, leads to important new insights. The identified variants, regions, and genes are candidates to modify disease presentation in cardiomyopathy.

PMID:40791945 | PMC:PMC12336239 | DOI:10.3389/fcvm.2025.1546493

BMC Med Inform Decis Mak. 2025 Aug 11;25(1):301. doi: 10.1186/s12911-025-03139-9.

ABSTRACT

BACKGROUND: Understanding early predictors of treatment outcomes allows better outcome prediction and resource allocation for efficient tuberculosis (TB) management.

OBJECTIVES: This study aimed to predict treatment outcomes of TB patients from a real-world population-wide health record dataset with a significant rate of incomplete observations. In addition, potential risk factors associated with death during TB treatment were investigated.

METHODS: We exploited the upweighting approach and multiple imputation analysis (MIA) to address the extreme imbalance in responses and missing data. Three algorithms were employed for TB treatment outcome prediction, including logistic regression (LOGIT), random forest, and stochastic gradient boosting. The three models exhibited similar performance in predicting the treatment outcomes. Moreover, an interpretation of LOGIT was conducted, adjusted odds ratios (aORs) were computed, and the interpretation results were compared between MIA and complete case analysis (CCA).

RESULTS: MIA was an appropriate method for coping with missing data. In addition, compared to CCA, the interpretation results of the MIA-derived LOGIT showed more statistically significant covariates associated with TB treatment outcomes. In MIA, factors such as TB clinical form involving both pulmonary TB and extrapulmonary TB [aOR = 3.077, 95% confidence interval (CI) = 2.994-3.163], retreatment after abandonment (aOR = 2.272, 95% CI = 2.209-2.338), and the absence of isoniazid (aOR = 2.072, 95% CI = 1.892-2.269) or rifampicin (aOR = 1.968, 95% CI = 1.746-2.218) in the treatment regimen were associated with increased odds of death.

CONCLUSION: In conclusion, our results shed light on the potential risk factors for death during TB treatment and suggest the use of simple yet interpretable LOGIT for the prediction of TB treatment outcomes.

PMID:40790736 | DOI:10.1186/s12911-025-03139-9

Mater Today Bio. 2025 Jul 28;34:102148. doi: 10.1016/j.mtbio.2025.102148. eCollection 2025 Oct.

ABSTRACT

Cystic fibrosis (CF) is an autosomal recessive disorder resulting from impaired anion transport in the epithelium of multiple organs, thereby affecting various physiological functions throughout the body. The heterogeneity of CF complicates drug development, highlighting the growing importance of individualized therapies. CF patient-derived organoid models and organ-on-a-chip (OOC) platforms are promising in vitro models for recapitulating CF pathology, owing to their high simulation fidelity, individualized therapeutic capabilities, cost-effectiveness, and high-throughput screening potential. This review systematically summarizes the technological development pathways of patient-derived organoids and OOC platforms for CF, along with recent advances in their applications to CF-related basic research, and particularly focuses on exploratory studies using organoid-on-a-chip (OrgOC) systems to elucidate CF pathogenesis and assess therapeutic approaches.

PMID:40791795 | PMC:PMC12336816 | DOI:10.1016/j.mtbio.2025.102148

Front Immunol. 2025 Jul 28;16:1595994. doi: 10.3389/fimmu.2025.1595994. eCollection 2025.

ABSTRACT

Cystic fibrosis (CF) lung disease manifests through abnormally thick mucus, persistent bacterial infections and a dysregulated innate immune system that involves significant neutrophilic inflammation. Neutrophils, immune cells essential to fight infections, accumulate in large numbers in CF airways and release neutrophil extracellular traps (NETs) into the airway lumen that deliver extracellular DNA, granule content and cytokines including IL-1β. Interleukin-1β, a powerful, proinflammatory cytokine, represents another, significant component of the innate immune system that is dysregulated in CF. Both defense mechanisms become problematic as NETs and IL-1β are present at elevated levels in CF airways, potentially creating a destructive cycle that exacerbates lung damage rather than protects against infections. Therefore, understanding the interplay between IL-1β and NETs is crucial for addressing CF lung disease progression. This review examines the general mechanisms of IL-1β release and NET formation, with particular focus on their role in CF lung disease, and proposes that a self-perpetuating, positive feedback loop between these two innate immune processes represents a major driving force in disease progression. This understanding suggests potential therapeutic targets for interrupting the cycle of inflammation and tissue damage in CF airways.

PMID:40791588 | PMC:PMC12337492 | DOI:10.3389/fimmu.2025.1595994

Front Cell Infect Microbiol. 2025 Jul 28;15:1602883. doi: 10.3389/fcimb.2025.1602883. eCollection 2025.

ABSTRACT

BACKGROUND: Non-puerperal mastitis (NPM) arises from heterogeneous factors ranging from autoimmune dysregulation to occult infections. To establish a diagnosis, biopsy is reliable but invasive. Imaging exhibits a limited specificity and may cause diagnostic delays, patient discomfort, and suboptimal management. Inspired by non-invasive tongue diagnosis in traditional Chinese medicine, this study integrated tongue-coating microbiota profiling and AI-quantified tongue image phenotyping to establish an objective, non-invasive diagnostic framework for NPM.

METHODS: A total of 100 NPM patients from the Breast Surgery Department of Longhua Hospital and 100 healthy volunteers were included. Their clinical characteristics, tongue images, and tongue-coating microbiota data were collected. Features of tongue images (detection, segmentation, and classification) were quantitated and extracted via deep learning. The microbiota composition was assessed using 16S rRNA gene sequencing (V3-V4 region) and bioinformatic pipelines (QIIME2, DADA2). Based on clinical, imaging, and microbial features, three machine learning models-logistic regression (LR), support vector machine (SVM), and gradient boosting decision tree (GBDT)-were trained to distinguish NPM.

RESULTS: The GBDT model achieved a superior diagnostic performance (AUROC = 0.98, accuracy = 0.95, and specificity = 0.95), outperforming the LR (AUROC = 0.98, accuracy = 0.95, and specificity = 0.90) and SVM models (AUROC = 0.87, accuracy = 0.80, and specificity = 0.75). Integration of clinical characteristics, tongue image features, and bacterial profiles (at the genus/family level) yielded the highest accuracy, whereas models using a single class of features showed a lower discriminatory ability (AUROC = 0.90-0.91). Key predictors included Campylobacter (12%), waist-hip ratio (11%), and Alloprevotella (6%).

CONCLUSIONS: Integrating clinical characteristics, tongue image features, and tongue-coating microbiota profiles, the multimodal GBDT model demonstrates a high diagnostic accuracy, supporting its utility for early screening and diagnosis of NPM.

PMID:40792098 | PMC:PMC12336138 | DOI:10.3389/fcimb.2025.1602883

Ophthalmol Sci. 2025 Jun 3;5(6):100837. doi: 10.1016/j.xops.2025.100837. eCollection 2025 Nov-Dec.

ABSTRACT

PURPOSE: Age-related cataract is the leading cause of vision impairment. Researchers have utilized various imaging modalities, including slit beam, diffuse anterior segment, and retinal imaging, to develop deep learning (DL) algorithms for automated cataract analysis. However, the comparative performance of these algorithms across different ocular imaging modalities remains unevaluated, mainly due to the absence of standardized test sets across studies.

DESIGN: Retrospective study.

PARTICIPANTS: Across all the models, the Singapore Malay Eye Study data set was used for training (N = 7093 eyes) and internal testing (N = 1649 eyes). The Singapore Indian Eye Study (SINDI; N = 5579 eyes) and the Singapore Chinese Eye Study (SCES; N = 5658 eyes) were used for external testing. A community study data set of nonmydriatic retinal photos (N = 310 eyes) was used for external testing of the retinal model.

METHODS: We developed 3 single-modality DL models (retinal, slit beam, and diffuse anterior segment photos) and 4 ensemble models (4 different combinations of the 3 single-modality models) to detect visually significant cataract (VSC). We defined eyes with VSC as having significant cataract (based on the modified Wisconsin cataract grading system) with a best-corrected visual acuity of <20/60.

MAIN OUTCOME MEASURES: Area under receiver operating characteristic curve (AUC).

RESULTS: In the internal test, the retinal model had the highest AUC value (97.0%; 95% confidence interval [CI], 95.9-98.2), compared with the slit beam model (AUC, 93.4%; 95% CI, 90.1-96.7; P diff = .029) and diffuse anterior segment model (AUC, 94.4; 95% CI, 92.3-96.4; P diff = .002). There was no significant difference in AUC when comparing the retinal model with the ensemble models (all P diff ≥ .07). These trends were consistently observed in the external test sets. In nonmydriatic eyes, the retinal model showed reasonable performance (AUC, 89.8%; 95% CI, 89.6-89.9).

CONCLUSIONS: Our findings highlight the retinal model as a promising tool for detecting VSC, outperforming slit beam and diffuse anterior segment models. Because retinal photography is routine in diabetic retinopathy screening, this approach could enable opportunistic cataract screening with minimal add-on cost.

FINANCIAL DISCLOSURE: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

PMID:40792062 | PMC:PMC12336814 | DOI:10.1016/j.xops.2025.100837

iScience. 2025 Jul 18;28(8):113158. doi: 10.1016/j.isci.2025.113158. eCollection 2025 Aug 15.

ABSTRACT

Detecting tumor-specific DNA methylation in circulating tumor DNA (ctDNA) offers a non-invasive method for tumor detection. The primary challenge lies in identifying the extremely low abundance of ctDNA in cell-free blood plasma (cfDNA). In this study, we present Oncoder, an interpretable deep learning-based tool for economical and accurate non-invasive tumor monitoring and diagnosis. Unlike other methods, Oncoder learns scientifically sound reference methylation atlases from patient blood to provide additional diagnostic insights, fostering trust among clinicians and patients, and continuously improves its accuracy through iterative learning. In simulations, Oncoder reduced prediction errors of tumor signals in blood by at least 30% compared to existing methods and showed the highest prediction correlation, indicating more accurate tumor progression monitoring. We also evaluated Oncoder's performance in various real-world applications. Oncoder sensitively detected changes in ctDNA levels during tumor development and treatment and exhibited superior diagnostic potential even in the earliest stages of cancer.

PMID:40792040 | PMC:PMC12337691 | DOI:10.1016/j.isci.2025.113158

Front Neurorobot. 2025 Jul 28;19:1604453. doi: 10.3389/fnbot.2025.1604453. eCollection 2025.

ABSTRACT

Myoelectric control systems translate electromyographic signals (EMG) from muscles into movement intentions, allowing control over various interfaces, such as prosthetics, wearable devices, and robotics. However, a major challenge lies in enhancing the system's ability to generalize, personalize, and adapt to the high variability of EMG signals. Artificial intelligence, particularly neural networks, has shown promising decoding performance when applied to large datasets. However, highly parameterized deep neural networks usually require extensive user-specific data with ground truth labels to learn individual unique EMG patterns. However, the characteristics of the EMG signal can change significantly over time, even for the same user, leading to performance degradation during extended use. In this work, we propose an innovative three-stage neural network training scheme designed to progressively develop an adaptive workflow, improving and maintaining the network performance on 28 subjects over 2 days. Experiments demonstrate the importance and necessity of each stage in the proposed framework.

PMID:40791943 | PMC:PMC12336220 | DOI:10.3389/fnbot.2025.1604453

Proc IEEE Int Symp Biomed Imaging. 2025 Apr;2025. doi: 10.1109/isbi60581.2025.10981108. Epub 2025 May 12.

ABSTRACT

With the rapid development of deep learning, CNN-based U-shaped networks have succeeded in medical image segmentation and are widely applied for various tasks. However, their limitations in capturing global features hinder their performance in complex segmentation tasks. The rise of Vision Transformer (ViT) has effectively compensated for this deficiency of CNNs and promoted the application of ViT-based U-networks in medical image segmentation. However, the high computational demands of ViT make it unsuitable for many medical devices and mobile platforms with limited resources, restricting its deployment on resource-constrained and edge devices. To address this, we propose EViT-UNet, an efficient ViT-based segmentation network that reduces computational complexity while maintaining accuracy, making it ideal for resource-constrained medical devices. EViT-UNet is built on a U-shaped architecture, comprising an encoder, decoder, bottleneck layer, and skip connections, combining convolutional operations with self-attention mechanisms to optimize efficiency. Experimental results demonstrate that EViT-UNet achieves high accuracy in medical image segmentation while significantly reducing computational complexity. The code is available at https://github.com/Retinal-Research/EVIT-UNET.

PMID:40791942 | PMC:PMC12337706 | DOI:10.1109/isbi60581.2025.10981108

Front Neurol. 2025 Jul 28;16:1615523. doi: 10.3389/fneur.2025.1615523. eCollection 2025.

ABSTRACT

Brain diseases pose a significant global health challenge due to their complexity and the limitations of traditional medical strategies. Recent advancements in artificial intelligence (AI), especially deep learning models like Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Graph Neural Networks (GNNs), offer powerful new tools for analysis. These neural networks are effective at extracting complex patterns from high-dimensional data. By integrating diverse data sources-such as neuroimaging, multi-omics, and clinical information-multimodal AI provides the comprehensive view needed to understand intricate disease mechanisms. This review outlines how these technologies enhance precision drug development and enable closed-loop treatment systems for brain disorders. Key applications include improving diagnostic accuracy, identifying novel biomarkers, accelerating drug discovery through target identification and virtual screening, and predicting patient-specific treatment responses. These AI-driven methods have the potential to shift medicine from a one-size-fits-all model to a personalized approach, with diagnostics and therapies tailored to individual profiles. However, realizing this potential requires addressing significant challenges related to data access, model interpretability, clinical validation, and practical integration.

PMID:40791911 | PMC:PMC12336123 | DOI:10.3389/fneur.2025.1615523

medRxiv [Preprint]. 2025 Jul 14:2025.07.11.25331386. doi: 10.1101/2025.07.11.25331386.

ABSTRACT

BACKGROUND: Determination of cardiac output (CO) is essential to the clinical management of cardiovascular compromise. However, the invasiveness, procedural risks, and reliance on specialized infrastructure limit accessibility and scalability of standard-of-care right heart catheterization (RHC). Seismocardiography (SCG), a non-invasive technique which records subtle chest wall vibrations generated by cardiac mechanical activity, may offer a promising alternative for CO determination.

OBJECTIVES: To develop and evaluate a deep learning model for estimating CO directly from SCG, electrocardiogram (ECG), and body mass index (BMI) in heart failure patients undergoing RHC.

METHODS: We trained a deep convolutional neural network for CO estimation using an open-access dataset comprising 73 heart failure patients with simultaneous RHC, SCG, and ECG recordings. Model performance was evaluated using a rotating leave-pair-out cross-validation strategy.

RESULTS: When estimating CO, the deep learning model achieved a mean bias of -0.35 L/min with limits of agreement (LoA) from -2.21 to 1.51 L/min. When predicting cardiac index in patients with a reference index < 2.2 L/min/m 2 , the model yielded a mean bias of 0.07 L/min/m 2 with LoA from -0.35 to 0.48 L/min/m 2 .

CONCLUSIONS: This study demonstrates the feasibility of using deep learning in combination with wearable SCG sensors to non-invasively estimate CO. Model performance was particularly strong in low-output states. These findings highlight the potential of SCG-based monitoring to augment clinical decision-making in settings where invasive measurements are impractical or unavailable. Prospective multicenter validation is needed to confirm generalizability and assess clinical impact.

SOURCES OF SUPPORT: This work was supported by NIH grants T32 HL129964 (N.J.K.), K08 ES037420 (N.J.K.), R01 HL124021 (S.Y.C.), R01 HL122596 (S.Y.C.); R01 HL151228 (S.Y.C.); the McKamish Family Foundation, the Hemophilia Center of Western Pennsylvania, and the Institute for Transfusion Medicine (N.J.K.), United Therapeutics Jenesis Innovative Research Award (N.J.K.), and the Pulmonary Hypertension Association (N.J.K.).

DISCLOSURES: S.Y.C. has served as a consultant for Merck, Janssen, and United Therapeutics; S.Y.C. is a director, officer, and shareholder in Synhale Therapeutics and Amlysion Therapeutics; S.Y.C. and N.J.K. hold research grants from United Therapeutics; S.Y.C. holds research grants from Bayer and the WoodNext Foundation. S.Y.C. has filed patent applications regarding the targeting of metabolism in pulmonary hypertension. Other authors: none.

TWITTER SUMMARY: We developed a deep learning model to non-invasively estimate cardiac output from wearable seismocardiogram (SCG) signals in patients undergoing right heart catheterization. This is the largest study to date using SCG for noninvasive cardiac output monitoring. #HeartFailure, #WearableTech, #AIInCardiology.

PMID:40791697 | PMC:PMC12338910 | DOI:10.1101/2025.07.11.25331386