Predicting the malignant transformation of rectal precancerous lesions remains challenging because conventional Whole Slide Images (WSIs) capture morphological information but lack molecular insight. Multiomics data provide complementary biological signals that often precede visible morphological changes. This study aimed to develop an artificial intelligence (AI)-based multimodal framework integrating WSI and multiomics data for accurate early prediction of malignant transformation.

WSI patches (512×512 px at 20× magnification) and matched multiomics profiles were used for 450 rectal tissue samples from the publicly available The Cancer Genome Atlas dataset. A multimodal architecture was designed, employing a Vision Transformer (ViT-B/16) for WSI feature extraction and a Variational Autoencoder for multiomics representation learning. Features were fused via a cross-attention mechanism to capture inter-modality dependencies. Baseline models, including a convolutional neural network-only image model and an omics-only multilayer perceptron, were trained for comparison. Five-fold cross-validation was applied, with binary cross-entropy loss, the AdamW optimizer, early stopping, and hyperparameter tuning to ensure reproducibility.

The multimodal Vision Transformer–Variational Autoencoder fusion model outperformed unimodal baselines, achieving an accuracy of 0.892 ± 0.012 and an area under the receiver operating characteristic curve of 0.927 ± 0.009, corresponding to a 7–10% improvement over WSI-only and omics-only models. Cross-attention–based fusion improved prediction stability and classification performance, while interpretability analyses (Grad-CAM and SHAP) highlighted biologically meaningful histopathological regions and molecular feature contributions.

This study presents a robust and scalable AI-based framework for integrating WSI and multiomics data in rectal precancerous lesions. The model improves predictive precision compared with unimodal baselines and offers preliminary interpretability insights through attention mechanisms. These findings support the potential of multimodal AI for early cancer risk assessment and precision pathology.

Colorectal cancer histopathological grading relies on the accurate segmentation of glandular structures. Current deep learning–based methods depend heavily on large-scale pixel-level annotations that are labor-intensive and not amenable to clinical practice. Weakly supervised semantic segmentation offers a promising alternative; yet, existing class activation map–based weakly supervised semantic segmentation approaches often produce incomplete, low-quality pseudo-masks that overemphasize discriminative regions and fail to provide reliable supervision for unannotated glandular structures, limiting their suitability for dense histopathology segmentation under sparse supervision. We propose a novel weakly supervised teacher–student framework that leverages sparse pathologists’ annotations and an Exponential Moving Average–stabilized teacher network to generate refined pseudo-masks.

Our framework integrates confidence-based filtering, adaptive fusion of teacher predictions with limited ground truth, and curriculum-guided refinement, enabling the student network to progressively delineate and accurately segment unannotated glandular regions. We validated our framework on an institutional colorectal cancer cohort from The Ohio State University Wexner Medical Center, consisting of 60 hematoxylin and eosin-stained whole-slide images from independent patients with varying degrees of gland differentiation, as well as on public benchmarks including the Gland Segmentation dataset (derived from stage T3–T4 colorectal adenocarcinomas), TCGA-COAD, TCGA-READ, and SPIDER.

The proposed framework achieved strong performance on the institutional dataset despite limited annotations. On the Gland Segmentation dataset, it demonstrated competitive performance compared to both weakly and fully supervised approaches, achieving a mean Intersection over Union of 80.10% ± 1.52 and a mean Dice coefficient of 89.10% ± 2.10. Moreover, cross-cohort evaluations showed robust generalization on TCGA-COAD and TCGA-READ without requiring additional annotations, while reduced performance on SPIDER reflected pronounced domain shift.

Our framework provides an annotation-efficient and generalizable paradigm for accurate gland segmentation in colorectal histopathology, offering a practical pathway toward significantly reducing annotation burdens while preserving high segmentation fidelity.

Metabolic dysfunction-associated fatty liver disease (MAFLD) has become one of the leading causes of chronic liver diseases in China, imposing a substantial and growing burden on the healthcare system. Considering the large number of individuals affected by MAFLD and the gap in disease management capacity at the primary care level, standardized guidance tailored to primary healthcare settings is urgently needed. In response, the Chronic Disease Management Branch of the China Medical Biotechnology Association convened a multidisciplinary working group incorporating hepatologists, general practitioners, and other specialists to initiate the first China national Guidelines for Diagnosis and Management of Metabolic Dysfunction-associated Fatty Liver Disease in Primary Care (2025). These guidelines provide recommendations and suggestions covering screening, risk assessment, diagnosis, treatment, referral pathways, and follow-up tailored for primary care institutions, thereby improving the long-term outcomes for the population with MAFLD and comprehensively strengthening the role of primary healthcare in chronic liver disease management.

Despite rapid advances in computational biology and regulatory reforms encouraging the reduction of animal use, a clear synthesis of how artificial intelligence (AI)-driven polypharmacology can function as a scientific and ethical bridge between traditional in vivo pharmacology and human-relevant drug development remains lacking. The shift from cage-based experimentation to code-based predictive modeling presents both opportunities and unresolved challenges in biological interpretation, regulatory acceptance, and pharmacology education. Therefore, this review aims to critically examine the transition toward AI-enabled, human-centric drug discovery within the framework of the 3R principles (Replacement, Reduction, and Refinement). Specifically, it explores (i) the global regulatory and ethical drivers accelerating non-animal methodologies, (ii) the scientific and educational gaps emerging from reduced dependence on animal models, and (iii) the role of AI and deep learning in reconstructing biological complexity through multi-omics integration and predictive toxicity modeling. By analyzing emerging AI platforms and computational strategies, this review highlights how AI-driven polypharmacology may offer a scalable, ethical, and precision-oriented framework for future pharmacological research.

Accurate identification of invasive fungal pathogens is crucial for appropriate antifungal therapy. The Department of Clinical Laboratory at Indus Hospital & Health Network, Karachi, Pakistan, reported two cases of invasive fungal infections between 1st January and 31st March 2024 in which conventional identification methods and automated systems produced discordant results, highlighting critical diagnostic challenges.

Two invasive yeast isolates initially showing budding yeast cells without pseudohyphae on Gram stain were subjected to conventional identification using cornmeal-Tween 80 agar, chrome agar, and BiGGY agar, followed by automated identification using the VITEK 2 ID-YST system and confirmatory API 20C AUX testing. Both isolates demonstrated typical soft, wrinkled, cream-colored colonies on Sabouraud dextrose agar, which on chrome agar appeared as dry, blue colonies and on BiGGY agar as dry, brown colonies. Characteristic arthroconidia and blastoconidia formation on cornmeal-Tween 80 agar were observed, consistent with Trichosporon species. However, the VITEK 2 ID-YST system identified both isolates as Cryptococcus laurentii with good confidence levels. India ink staining was negative for both isolates. Confirmatory API 20C AUX testing correctly identified both isolates as Trichosporon asahii (identification profile 3740734).

This discordance between automated and conventional methods underscores the continued importance of conventional identification techniques and highlights potential limitations of automated systems for certain uncommon yeasts. Laboratories should maintain proficiency in conventional methods and consider confirmatory testing when automated results conflict with morphological findings. The clinical implications of misidentification include inappropriate antifungal selection, given the different susceptibility patterns between these species.

Artificial intelligence (AI) translation in genitourinary (GU) pathology has progressed unevenly across organs and tasks. This review addresses a central clinical question: which GU pathology AI applications are deployment-ready, which require further validation, and what frameworks can guide safe implementation? We synthesize evidence across GU organs and introduce pragmatic translation frameworks to guide deployment and prioritize translational research.

Narrative review integrating foundational literature with targeted 2023–2025 publications, emphasizing regulatory milestones, external validation, and prospective studies. Literature was identified through PubMed, Embase, and conference proceedings using structured search terms for AI, digital pathology, and GU organ-specific queries. For each organ/task, we mapped evidence strength, regulatory maturity, generalizability, workflow integration, safety, and feasibility to a Translational Readiness Index (TRI) rubric (0–30 scale).

Prostate biopsy AI demonstrates the strongest maturity (TRI 26/30), supported by U.S. Food and Drug Administration-cleared systems, multi-site validation, and prospective implementations showing efficiency gains and reduced ancillary testing. Bladder cytology shows moderate readiness (TRI 19/30), with commercial offerings supporting pilotable prescreening workflows aligned with the Paris System when paired with uncertainty-aware deferral. Bladder histology, renal neoplasia, and low-prevalence domains (testis, penis) remain emerging (TRI 6–15/30), constrained by label variability, rare subtype underrepresentation, and limited external validation.

The TRI rubric, SURE-Path safety bundle, and VALIDATED/ORCHESTRATE implementation pathway provide a practical template for evidence-based deployment in GU pathology. Clinically defensible translation requires matching intended use to validation evidence, with explicit safeguards for emerging applications.

Fast inverse planning in radiosurgery planning is limited by an excessive number of isocenters, which is clinically hypothesized to be driven by the morphological irregularity of the target volume. This retrospective cross-sectional study aimed to empirically evaluate this hypothesis in vestibular schwannoma cases.

Consecutive patients diagnosed with vestibular schwannoma and receiving Gamma Knife radiosurgery in 2023 were included, and their treatment plans designed using the GammaPlan planning system were collected. Morphological irregularity–related parameters, including standard sphericity (SS), volume ratio sphericity (VRS), and the coefficient of variance of diameters (DCV), were calculated based on parameters provided by the system. Basic demographic and clinical data were collected to evaluate their impact on sphericity. The effects of different sphericity assessment methods on common treatment plan parameters were analyzed.

Treatment plans of 280 patients with vestibular schwannoma were collected. The SS, VRS, and DCV of the tumors were 0.85 (0.77–0.91), 0.46 ± 0.16, and 0.22 (0.14–0.34), respectively. Multivariate analysis showed that lesion volume, acoustic neuroma consensus on systems for reporting results grade, and age were significant factors influencing sphericity. All other planning parameters, except prescription dose and homogeneity index, were significantly correlated with sphericity. DCV was more closely correlated with SS than with VRS.

DCV may serve as a simple quantitative metric of target morphological irregularity, showing strong consistency with SS. Incorporating morphological irregularity into Gamma Knife treatment plan evaluation may help improve future planning strategies and support optimization of isocenter utilization.

Chronic pancreatitis is an inflammatory disease and is difficult to manage despite advancements in medical science. This study examined the effect of water/ethanol extracts of Justicia carnea leaves on oxidative stress and glucagon expression in a mouse model of chronic pancreatitis induced by trinitrobenzenesulfonic acid (TNBS).

Twenty-five male Swiss albino mice were randomized and treated intrarectally with vehicle (the control group) or TNBS. Some TNBS-treated mice were treated orally with 200 mg/kg or 400 mg/kg J. carnea extracts, or with the positive control, 500 mg/kg sulfasalazine, every other day on three occasions. Oxidative stress markers and pancreatic glucagon expression were assessed.

Compared with the healthy control mice, treatment with TNBS significantly decreased the levels of pancreatic glutathione (0.89 µmol/g tissue vs. 7.16 µmol/g tissue in the control) and glutathione peroxidase activity, but significantly increased the levels of α-amylase and lipase activities, lipid peroxidation, total antioxidant capacity, and nitric oxide, as well as serum C-reactive protein (P < 0.05 for all), accompanied by severe inflammation and reduced glucagon expression in the pancreatic tissues. The toxic effects of TNBS were significantly mitigated by treatment with J. carnea extracts.

These findings provide evidence that treatment with J. carnea extracts inhibited oxidative stress and preserved glucagon expression in the pancreatic tissues of mice.

Given the lack of efficient biomarkers for hepatocellular carcinoma (HCC) diagnosis, this study aimed to develop an HCC diagnostic strategy based on serum protein glycosylation signatures. We characterized differential N-glycosylation patterns of serum IgG to differentiate HCC from healthy controls and liver cirrhosis, and elucidated the molecular mechanisms driving aberrant Neu5Gc elevation in HCC to provide a theoretical basis for clinical application and differential diagnosis of HCC.

LIP-ELISA was applied to quantify serum Neu5Gc in 6,768 healthy individuals for baseline establishment. IgG was purified and subsequently analyzed by RPLC-MS/MS for glycosylation profiling in HCC and healthy samples. Bioinformatic analysis of CMAH and related gene clusters modulating Neu5Gc synthesis was conducted.

In a cohort of 1,114 participants, the LIP-ELISA platform achieved 80.21% sensitivity, 96.01% specificity, and 92.46% accuracy for primary HCC diagnosis. Serum IgG from HCC patients displayed multi-branched N-glycans modified with core fucose and Neu5Gc. Key molecules involved in glycan modification were identified, enabling the development of multiplexed gene detection for HCC, LC, and chronic hepatitis B. In vitro assays confirmed hypoxia-induced sialic acid accumulation in HCC cells. Meanwhile, CMAH-knockout mouse experiments verified that an exogenous high-sialic-acid diet compensates for endogenous Neu5Gc synthesis deficiency, revealing a dietary-mediated compensatory mechanism for Neu5Gc elevation.

This study established an LIP-ELISA-based clinical diagnostic platform combining AFP and Neu5Gc, defined sialic acid–modified glycan structures, and preliminarily identified regulators of Neu5Gc biosynthesis, providing novel insights for HCC diagnosis and mechanism research.