Find the papers that actually matter

To test whether the mean curvature of isophotes (MCI), a geometric image transformation, can be used to improve automatic detection on chest CT of Usual Interstitial Pneumonia (UIP), a determining radiological pattern in the diagnosis of Interstitial Lung Diseases (ILD). This retrospective study included chest CT scans from 234 patients (123 female,111 male; mean age: 61.6 years; age range: 18-90 years) obtained at two independent institutions between 2007 and 2024.Three different classification models were trained on the original CT images and separately on MCI-transformed CT images: (1) a previously published deep learning model for classifying fibrotic lung disease on chest CT, (2) a classification pipeline based on the EfficientNet-V2 convolutional neural network architecture, and (3) a non-deep-learning model based on the functional principal component analysis (FPCA) of density functions of voxel intensity.All models were trained on data from the first institution and evaluated on data from the second institution with the recall-macro, precision-macro and F1-macro scores. Performance difference between classifier pairs was tested with the Stuart-Maxwell marginal homogeneity test. For a fixed model architecture and training algorithm, MCI-transformed images yield comparable or better classification performance than the original CT images. The best performance improvement achieved with MCI compared to CT was: recall-macro 0.83 vs 0.57, precision-macro 0.81 vs 0.50, F1-macro 0.80 vs 0.49, p = 4.2e-5. MCI may be a valuable addition to existing AI systems for screening for UIP on chest CT.

Compared to non-triple-negative breast cancer (Non-TNBC), triple-negative breast cancer (TNBC) exhibits significantly poorer prognosis. Previous research has confirmed that the PI3K/AKT pathway is closely associated with prognosis in breast cancer patients. Yet, it remains unclear whether this pathway is implicated in the prognostic differences observed between TNBC and Non-TNBC. After downloading raw transcriptomic datasets from the GEO database and removing batch effects, we performed an integrated analysis to delineate how key genes drive the poor prognosis of TNBC. Functional enrichment, machine-learning-based feature selection, immune-cell infiltration profiling, drug-sensitivity screening, single-cell RNA sequencing and spatial transcriptomics were successively applied. Molecular-docking simulations were finally conducted to evaluate the binding affinity of MYB toward bioactive compounds derived from the Taohong Siwu Decoction. Across 113 algorithm combinations, MYB plays the most critical role in distinguishing TNBC from Non-TNBC. The constructed prognostic model confirms the significant association between MYB expression and patient outcomes. Immune cell infiltration, drug sensitivity, single-cell data analysis and spatial transcriptome revealed the specific mechanisms through which MYB influences patient prognosis. Molecular docking experiments demonstrate strong binding between key components in Taohong Siwu Decoction and MYB. Based on multi-omics analysis, our findings indicate that the PI3K/AKT pathway is a key factor contributing to the significant prognostic disparity between TNBC and Non-TNBC. Within this pathway, the MYB gene emerges as a potential therapeutic target. This discovery provides a potential basis for future research exploring MYB as a therapeutic target for TNBC patients.

Extracellular vesicle-derived miRNAs (EV-miRNAs) have emerged as pivotal liquid biopsy biomarkers due to their biogenesis-driven selective enrichment, exceptional stability within lipid bilayers, and essential regulatory roles in various diseases. This review provides a systematic and critical evaluation of the field by integrating molecular mechanisms, high-performance sensing technologies, and clinical translation progress, underscoring the necessity of accurate EV-miRNA quantification for disease diagnosis, treatment, and prognosis monitoring. Various emerging sensing technologies have been developed to enhance detection sensitivity and specificity, such as enzyme-based amplification strategies, membrane fusion-mediated approaches, nanomaterial-assisted signal conversion techniques, and integrated microfluidic platforms. Furthermore, the review summarises representative clinical advances in using EV-miRNAs for the diagnosis and monitoring of cancers, neurodegenerative disorders, and metabolic diseases. Significant emphasis is also placed on addressing analytical limitations, with the integration of AI-driven multi-omics models highlighted as a strategic frontier for enhancing clinical utility. This comprehensive synthesis establishes a robust theoretical and technical foundation for advancing EV-miRNA research from laboratory validation toward high-precision clinical translation.

This study employed an integrative computational and systems biology framework to define a diagnostic gene signature for hepatocellular carcinoma (HCC) and to explore its potential translational relevance in a hypothesis-generating manner. Differential expression analysis of transcriptomic data from 230 samples identified 2748 significantly differentially expressed genes (DEGs), including 2283 upregulated and 465 downregulated genes, with FGF4 (log2FC = 10.08) and REG1B (log2FC = 10.02) among the top hits. Four machine learning classifiers were trained using this signature and demonstrated consistently high predictive performance, with XGBoost emerging as the top-performing model (accuracy = 0.97, F1-score = 0.96, ROC-AUC = 0.981). Logistic Regression (L1) and Random Forest achieved comparable performance (ROC-AUC = 0.980 and 0.979, respectively), while SVM-linear also showed high robustness (ROC-AUC = 0.978). All models showed good calibration, with low Brier scores (<0.04) and precision consistently exceeding 0.90 across most recall thresholds, indicating strong but not perfect classification performance. SHAP-based explainability analysis was used to rank and prioritise the most influential predictors, refining the biomarker panel to 81 genes that collectively accounted for approximately 50 % of the model's explanatory contribution, and highlighting key downregulated predictors in HCC, including GDF2, COLEC10, BMP10, LRAT, and DNASE1L3. Protein-protein interaction and functional enrichment analyses revealed five major molecular clusters and provided systems-level insights into dysregulated biological processes associated with HCC. Drug-gene interaction mining mapped 78 target proteins to clinically relevant compounds, including tolrestat, alcuronium, metyrosine, and 4-phenylbutyric acid. Molecular docking suggested favorable binding propensities for several complexes, including alcuronium-3UON (-8.5 kcal/mol), tolrestat-1ZUA (-8.3 kcal/mol), metyrosine-2XSN (-6.7 kcal/mol), and 4-phenylbutyric acid-2NZ2 (-5.9 kcal/mol). A 100 ns molecular dynamics simulation of the tolrestat-AKR1B10 (1ZUA) complex indicated structural stability, with protein backbone RMSD stabilising at 1.5-3.0 Å, ligand RMSD at 0.6-1.4 Å, and persistent interactions involving Trp22, His110, Glu111, and Phe122. Physicochemical and pharmacokinetic profiling further prioritised tolrestat as a computationally favourable candidate (MW = 357.35, LogP = 3.64, TPSA = 81.86 Ų), exhibiting acceptable drug-likeness, high predicted gastrointestinal absorption, and low synthetic complexity (SA = 2.34), in contrast to alcuronium (MW = 666.89, SA = 7.86), which showed multiple rule violations. Collectively, this in silico study proposes a robust diagnostic gene signature for HCC and identifies tolrestat as a promising repurposing candidate that warrants experimental validation, demonstrating the utility of integrating machine learning, network biology, and molecular simulation in translational cancer research.

Approximately 30-50% of Papillary thyroid carcinoma (PTC) patients develop cervical lymph nodes (LNs) metastasis, significantly increasing the risk of disease recurrence and impacting long-term outcomes. We introduced an open-access multicenter lymph node ultrasound image database (LymphUs) specifically designed to advance research in LN assessment for PTC. Ultrasound imaging was performed on PTC patients at two independent clinical centers using standardized acquisition protocols. Experienced radiologists at each center documented sixteen semantic features for each LN. All LNs were annotated with segmentation masks serving as ground truth, and classification into benign or malignant categories was confirmed by fine needle aspiration biopsy results. The LymphUs comprises ultrasound images with segmentation masks from 338 PTC patients with suspected LN metastasis, divided into two center-specific cohorts: 180 patients (81 malignant, 99 benign) and 158 patients (82 malignant, 76 benign). The complete dataset, including semantic features and expert annotations, is freely accessible for research purposes. The LymphUs bridges a critical gap in medical imaging resources by providing a large-scale, multicenter ultrasound database for cervical LN assessment in PTC, supporting diagnostic algorithms, standardized reporting systems, and artificial intelligence applications to enhance preoperative LN staging and treatment planning.

The response of resectable non-small cell lung cancer (NSCLC) to neoadjuvant immunotherapy is heterogeneous. Machine learning can integrate multimodal data to construct predictive models, but the methodological quality, risk of bias and clinical applicability of such models have not been systematically evaluated. This study aims to systematically evaluate the methodological quality, risk of bias, and diagnostic performance of machine learning models for predicting neoadjuvant immunotherapy response in resectable NSCLC. As of August 22, 2025, 11 databases were retrieved. Two researchers independently extracted the data, and a third researcher resolved the data differences. The quality of the model, the development process and the quality of radiomics reports were evaluated respectively by probast + AI, IJMEDI checklist and RQS. Meta-analysis of the AUC, sensitivity and specificity of the model was conducted using R software, and subgroup analysis was performed according to predictors, algorithms and outcomes. Seventeen studies involving 44 models were included. Eighty-nine percent of models had relatively low quality and all had a high risk of bias - key flaws included unreasonable sample size, improper handling of missing data and defects in validation procedures - but the overall applicability was good. IJMEDI scores ranged 26.5-37 (4 high-quality, others medium); average RQS of 12 radiomics studies was 14.58 (22.22%-52.78%), with multiple deficiencies. Ten internal validation models showed that the combined internal AUC was 0.786 (95% CI: 0.740-0.826, I2 = 0%), there was no publication bias (Egger's test), and the sensitivity was 0.763 (95% CI: (0.56-0.89), with a specificity of 0.908 (95% CI: 0.471-0.991). The predicted AUCs of MPR and PCR were 0.805 and 0.761, respectively. SVM achieved the highest AUC (0.841), and the non-radiomics model (0.869) was superior to the radiomics model (0.775). The combined external validation AUC was 0.760, among which the AUC predicted by MPR was 0.754. ML models show potential for predicting neoadjuvant immunotherapy efficacy in resectable NSCLC, with SVM and non-radiomics models superior. However, low methodological quality and high bias risk require cautious interpretation. Future work should refine methodology, address radiomics gaps, and promote clinical translation.

Preoperative imaging prediction of perineural invasion in gastric cancer (GC-PNI) mainly relies on tumour characteristics and clinical variables, while the potential of non-tumour-derived multimodal features remains underexplored. We retrospectively enrolled 777 patients from three medical centers and divided them into a training cohort, an internal testing cohort (I-T), and two external testing cohorts (E-T1, E-T2). We developed an end-to-end multimodal and multitask deep learning framework, termed GAST-NET, that integrates tumour CT features, visceral adipose tissue characteristics, and clinical variables to jointly predict perineural invasion (PNI) and five-year prognostic survival risk (PR). The model incorporates an Adaptive Multi-scale Feature Fusion Module (AMFM) and a Cross-Scale Fusion Pooling (CSF Pooling) module to capture hierarchical semantic information and enhance discriminative cross-modal representation. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), and decision curve analysis (DCA). Furthermore, five radiologists were invited to participate in the image reading experiment to verify the clinical interpretability and diagnostic gain of the model. The proposed model achieved AUCs of 0.923 (95% CI: 0.865-0.969), 0.868 (95% CI: 0.791-0.934), and 0.871 (95% CI: 0.806-0.930) for PNI prediction across the internal and two external cohorts, respectively. For prognostic risk prediction, the AUC reached 0.873 (95% CI: 0.835-0.922). When used as a decision-support tool, GAST-NET significantly improved diagnostic accuracy and reduced misclassification compared with radiologists. GAST-NET demonstrated strong generalizability and potential clinical utility in predicting perineural invasion (PNI) and prognosis in gastric cancer. Notably, visceral adipose tissue features provided complementary value for PNI prediction beyond conventional tumour characteristics.

Malignant tumors present a significant global health challenge, and accurate pathological grading is essential for personalized treatment. Traditional grading methods, which rely on invasive biopsies, are limited by tumor location. In contrast, magnetic resonance imaging (MRI) offers a non-invasive, high-resolution tool with multi-sequence MRI (e.g., T1, T2, T1C) enabling comprehensive tumor assessment. However, existing methods often struggle to capture cross-modal correlations and global dependencies. To address this limitation, we propose the Multimodal Hybrid Mamba (MSHM) classification model for tumor pathological grade prediction. The model integrates convolutional neural networks for shallow feature extraction, Mamba encoders for modeling global dependencies, and cross-modal attention to fuse multi-sequence MRI data. The Mamba-Fusion module further refines the global features, enhancing lesion recognition and computational efficiency. Experimental results demonstrate that MSHM outperforms existing methods, achieving 98.36 ± 1.00% AUC and 92.08 ± 3.26% F1-Score on the private orbital adnexal lymphoma dataset from multi-centers, and 98.93 ± 0.19% AUC and 95.82 ± 0.62% F1-Score on the public glioma BraTS 2024 dataset. Additionally, MSHM performs exceptionally well on the LLD-MMRI dataset, achieving 99.25 ± 0.26% AUC and 96.97 ± 0.55% F1-Score in distinguishing between benign and malignant liver lesions, further validating the model's robust performance across diverse datasets. Ablation studies confirm the effectiveness of the proposed modules. Overall, MSHM strikes a balance between high performance and efficiency, advancing both tumor pathological grade prediction and multimodal medical image analysis.

Melanoma is a highly aggressive form of skin cancer with poor prognosis. CD47, a well-known "don't eat me" signal, plays an important role in tumor immune evasion. In addition to immune checkpoint blockade, the STING-related pathway has emerged as a critical mechanism in activating innate immunity against tumors. Importantly, the antitumor effects of CD47 blockade require STING. To address this, we developed a novel "self" signal-suppressed Metal-Organic Framework (MOF) nanodrug, functionalized with anti-CD47 antibodies and loaded with the STING agonist diABZI (di@MOF@A). The "self" signal-suppressed MOF nanodrug effectively inhibits melanoma cell proliferation and induces significant ROS generation. The di@MOF@A achieves tumor targeting, and successfully blocks the CD47-SIRPα interaction, therefore reducing "Don`t eat me" signals. The nanodrug was designed to enhance chemodynamic therapy and immunotherapy effectiveness through a dual strategy of CD47 blockade and STING activation. The tumor mouse models displayed superior tumor-targeting ability of di@MOF@A, leading to substantial tumor growth inhibition (P < 0.01) and prolonged survival (P = 0.0003), without inducing significant systemic toxicity. These findings suggest that the MOF-based nanodrug represents a promising strategy for melanoma treatment with immune modulation to enhance therapeutic efficacy. Importantly, the successful integration of CD47 blockade and STING activation within a single nanoplatform offers significant potential for clinical translation, paving the way for developing MOF-based nanomedicines with CD47 blockade for enhanced immunotherapy.

In this study, we developed BioFluxNet, a 1D CNN-based algorithm for automated analysis of raw electrical signals in impedance cytometry to directly classify cell types and quantify cell counts. The network comprises three functional blocks: a feature extraction module with five alternating convolutional, normalization, activation, and pooling layers; a classification block with a single linear layer; and a counting block incorporating two linear layers and an activation layer. Raw signal streams of particles and cells, collected from an impedance cytometry featuring a serpentine microchannel and four pairs of face-to-face electrodes, are stored as training and testing sets, and then used to train and test BioFluxNet. Results demonstrate that the well-trained network achieves robust classification of raw signal streams from diverse particles and cells (including blood and tumor cells), while simultaneously enabling accurate counting of particles or cells form corresponding signal streams. Compared to conventional signal processing methods, BioFluxNet eliminates many time-consuming signal processing steps, reduces manual intervention, and minimizes subjectivity of operators. The proposed deep learning framework offers a rapid, automated solution for electrical signal analysis in impedance cytometry, showcasing broad applicability in cell characterization and related biomedical fields.