Although Vision Language Models (VLMs) have shown generalization in medical imaging, pathology presents unique challenges due to ultra-high resolution, complex tissue structures, and nuanced semantics. These factors make pathology VLMs prone to hallucinations, i.e., generating outputs inconsistent with visual evidence, which undermines clinical trust. Existing RAG approaches in this domain largely depend on text-based knowledge bases, limiting their ability to leverage diagnostic visual cues. To address this, we propose Patho-AgenticRAG, a multimodal RAG framework with a database built on page-level embeddings from authoritative pathology textbooks. Unlike traditional text-only retrieval systems, it supports joint text–image search, enabling retrieval of textbook pages that contain both the queried text and relevant visual cues, thus avoiding the loss of critical image-based information. Patho-AgenticRAG also supports reasoning, task decomposition, and multi-turn search interactions, improving accuracy in complex diagnostic scenarios. Experiments show that Patho-AgenticRAG significantly outperforms existing multimodal models in complex pathology tasks like multiple-choice diagnosis and visual question answering.

BACKGROUND The incidence of reflux esophagitis (RE) is increasing worldwide. RE can cause heartburn, acid reflux, and swallowing discomfort, thereby diminishing quality of life, with potential progression to Barrett's esophagus or esophageal adenocarcinoma. Gastroscopy for the diagnosis and grading of RE can be uncomfortable, and other modalities lack predictive accuracy. Thus, highly non-invasive diagnostic methods for RE are needed. AIM To explore the value of contrast-enhanced ultrasound (CEUS) examination for predicting the progression of RE. METHODS We retrospectively selected 296 patients with RE who were hospitalized at West China Longquanyi Hospital of Sichuan University from February 2019 to January 2023. Basic clinical characteristics such as age and gender were collected. CEUS was used to measure the diameter of the esophageal hiatus, frequency of reflux within 5 min, duration of reflux within 5 min, and length of the abdominal esophagus. According to the Los Angeles (LA) classification, the patients were divided into three groups: 92 patients with “LA-A” grade (Group A), 85 with “LA-B” grade (Group B), and 20 with “LA-C/D” grade (Group C). Group differences in clinical characteristics were analyzed. Key factors influencing the progression of RE were analyzed using ordered logistic regression, random forest, and support vector machine (SVM). Statistical methods included paired the t -test, Mann–Whitney U test, Chi-square test, Kruskal–Wallis H test, and one-way analysis of variance. RESULTS Comparisons among the three groups showed statistically significant differences in the diameter of the esophageal hiatus and frequency and duration of reflux within 5 min. Pairwise comparisons revealed that the frequency of reflux within 5 min in group C was significantly higher than in groups A and B. Ordered logistic regression, random forest, and SVM analyses all indicated that the frequency and duration of reflux within 5 min were key factors influencing the progression of RE. The SVM analysis showed that the diameter of the esophageal hiatus was a key factor influencing the progression of RE. CONCLUSION CEUS can better predict RE progression. The number of reflux events within 5 min, duration of reflux within 5 min, and diameter of the esophageal hiatus are key factors influencing RE progression.

In the occurrence and progression of breast cancer, tumor angiogenesis and metastasis play a central role. This indicates that anti-angiogenesis holds a significant position in anti-tumor therapy. Demethoxycurcumin is a bioactive diarylheptanoid compound extracted from Curcuma longa, a herb that serves both as food and medicine. As a member of the curcuminoid family, it exhibits anti-angiogenic and anti-tumor characteristics. But, the understanding of its potential mechanisms of action remains incomplete. Our experiments demonstrated that demethoxycurcumin significantly inhibited tumor growth (Ki67) and microvascular density (CD31) in the 4T1 breast cancer mouse model. Our in vitro experiments revealed that demethoxycurcumin inhibits the proliferation, migration, and angiogenesis of HUVECs in a dose-dependent manner. The underlying mechanism is characterized by a reduction in RGS5 expression in endothelial cells that proliferate as a consequence of tumor pathological features. The efficacy of high doses of demethoxycurcumin could be reversed by overexpression of RGS5, indicating that the effect of demethoxycurcumin is mediated through RGS5. These findings confirm that RGS5 is a key target for demethoxycurcumin in inhibiting angiogenesis in breast cancer, elucidating its anti-tumor mechanisms and providing new references for future research.

Sentinel lymph node (SLN) biopsy remains the standard for axillary staging in early-stage breast cancer, though ongoing clinical investigations are exploring the omission of axillary procedures in specific subgroups. This study assessed whether axillary ultrasonography (US) and MRI can predict SLN involvement and developed a predictive tool to identify patients who may safely forgo axillary surgery. We retrospectively analyzed 8114 patients with cT1–T2N0 invasive breast cancer across three cancer centers in China. All patients underwent preoperative axillary US and/or MRI. Multivariate logistic regression identified independent predictors of SLN metastases, which were used to construct a predictive model. The model was validated using a 70:30 training-validation split and visualized through a nomogram. Subgroup analyses evaluated the risk of SLN involvement among patients with negative imaging findings. SLN metastases were observed in 2545 patients (31.37

The development and clinical application of novel HER2-targeted antibody-drug conjugates (ADCs) have significantly improved outcomes for breast cancer patients with HER2-low expression. The efficacy of such therapies critically depends on accurate assessment of HER2-low status. However, immunohistochemical (IHC) interpretation of HER2-low breast cancer faces multiple challenges due to tumor heterogeneity and interobserver variability among pathologists. This study aimed to develop a deep learning-based framework for analyzing hematoxylin and eosin (H E) stained whole-slide images (WSIs) of breast cancer to achieve precise prediction of HER2-low status while providing interpretable evidence. We retrospectively collected 776 cases of invasive breast carcinoma diagnosed at the Affiliated Hospital of Zunyi Medical University between January 2019 and April 2023 to construct a HER2-low expression dataset. Leveraging an ImageNet-pretrained ResNet50 model for feature extraction and a CLAM (Clustering-constrained Attention Multiple Instance Learning) model with 10-fold cross-validation, our framework demonstrated robust performance on both validation and test sets. Critical HER2-low predictive regions were visualized using attention heatmaps to enhance model interpretability. The mean AUC of the model was 0.613 ± 0.118 on the validation set, and 0.608 ± 0.104 on the test set. The attention heatmap visualization provided biologically plausible explanations for model predictions, offering reliable decision-support tools for pathologists.

The immune defense function protecting the body from invasive pathogens is a key indicator of an individual's health and lacks of methods for quantitative evaluation. This study introduces ImmuDef, a novel algorithm for precisely and quantitatively assessing anti-infection immune defense function based on RNA-seq data. ImmuDef selects immune signatures through comparisons of acquired immunodeficiency syndrome (AIDS) or severe sepsis vs. healthy controls (HC) and reduces dimension to construct a latent space via a variational autoencoder (VAE) model (QImmuDef-VAE), a representation deep learning model. Based on this model, a defense immune score (DImmuScore) was calculated by measuring the distance between a patient and HC within latent space. We validated ImmuDef on 3202 samples across four immune states: immunodeficiency, immunocompromised, immunocompetent, and immunoactive. As a result, DImmuScore achieves high classification accuracy (mean accuracy: 71.75%-76.25%) among samples with various immune states and infections. Furthermore, DImmuScore can serve as a metric for infectious disease severity, where its gradient directly quantifies disease severity. As an application, DImmuScore can be a strong prognostic indicator, effectively stratifying mortality/survival in both sepsis and COVID-19 patients with no symptomatic difference. This framework was validated across five infectious diseases, establishing the first quantitative standard for cross-disease immune defense assessment.

Breast core needle biopsy (CNB) is central to breast cancer diagnosis yet remains challenging because limited tissue sampling, lesion heterogeneity, and subtle morphologic overlap can obscure subtype distinctions. We developed CorePath, a breast-specialized multimodal pathology foundation model fine-tuned from PRISM using 7901 paired CNB whole-slide images and diagnostic reports from two centers. Evaluated across six CNB cohorts and two public breast pathology benchmarks without task-specific retraining, CorePath consistently outperformed PRISM across cancer detection, invasion assessment, and histological subtyping. It achieved weighted area under the receiver operating characteristic curves (AUCs) of 0.9526-0.9735 for five-class CNB histological subtyping across private centers. On public benchmarks, CorePath outperformed leading pathology foundation models, achieving the highest weighted AUCs of 0.7780 for BCNB invasive carcinoma subtyping, 0.8178 for BRACS lesion stratification, and 0.8252 for BRACS fine-grained classification. In report generation, CorePath reduced the overall non-breast hallucinations from 30.1

Objective To investigate the expression pattern,prognostic significance,and underlying molecular mechanisms of mutY homolog(MUTYH)in hepatocellular carcinoma(HCC),and to evaluate its clinical potential as a novel biomarker and therapeutic target.Methods The differential expression of MUTYH between HCC and normal tissues was compared using the TCGA and GEO databases.Associations with clinicopathological parameters,TP53 mutation status,diagnostic efficacy of alpha-fetoprotein(AFP),and sorafenib resistance were analyzed.Prognostic impact was evaluated using the Kaplan-Meier method with the log-rank test,and univariate and multivariate Cox proportional hazards regression models were used to verify its independent prognostic value.Molecular mechanisms were explored through GO,KEGG,and GSEA enrichment analyses as well as protein-protein interaction(PPI)network construction.The correlations of MUTYH with immune cell infiltration and the immunotherapeutic efficacy of immune checkpoint inhibitors were assessed using the CIBERSORT algorithm and the BEST database.Quantitative real-time PCR(qPCR)was performed to validate the expression differences of MUTYH and its core interacting molecules between HCC and normal tissues.Results MUTYH was significantly upregulated in HCC(P<0.05),clinical sample tests have confirmed that it can serve as a biomarker for diagnosing HCC(area under the curve[AUC]=0.824,95％CI:0.762-0.886,P<0.001),its diagnostic value remains high even in the HCC subgroup with low AFP expression(GSE25097,AUC=0.716,P<0.001;GSE63898,AUC=0.624,P<0.001).High MUTYH expression correlated with sorafenib resistance(P<0.05)and was an independent risk factor for poor overall survival(hazard ratio[HR]=1.92,P<0.05).Mechanistically,MUTYH was positively associated with apurinic/apyrimidinic endonuclease 1(APEX1)(r=0.83,P<0.05),potentially facilitating G1/S transition by modulating cyclin-dependent kinase 4,cyclin-dependent kinase 7,and cyclin E2.Immune analysis identified MUTYH as a predictor for anti-PD-1/PD-L1 response(IMvigor210 AUC=0.637;Cho2020 AUC=0.782),though no association was found with anti-CTLA-4 therapy.Conclusion MUTYH is significantly overexpressed in HCC and may promote HCC progression by regulating APEX1 and key cell cycle molecules.Compared with the conventional marker AFP,MUTYH demonstrates superior diagnostic and prognostic evaluation efficacy and is associated with anti-PD-1/PD-L1 therapeutic response and sorafenib resistance.Overall,MUTYH has potential as a novel biomarker and therapeutic target for HCC.

ABSTRACT Migrasomes, newly discovered vesicular organelles, hold promise as diagnostic biomarkers and therapeutic targets in various diseases. However, the exploration of their clinical value remains hindered by the complexity of enriching and analyzing low concentrations of migrasomes in body fluids. To address this issue, a magnetic‐assisted strategy was devised for screening aptamers specific to migrasomes, with the identified aptamer then being utilized for the specific isolation of migrasomes derived from clinical plasma. Initially, lipid‐affinity magnetic nanoparticles were prepared and employed in a Magnetic‐Systematic Evolution of Ligands by Exponential Enrichment (Mag‐SELEX) process to identify aptamers that specifically target migrasomes. An optimal aptamer, Apt_B3, with a dissociation constant (Kd) of 251.9 nM, was successfully identified. This aptamer was subsequently utilized to construct the magnetic aptamer probe system, enabling the precise and rapid capture of migrasomes from plasma within 15 min. Our strategy exhibited exceptional separation efficiency, confirming its reliability and enhanced performance compared to traditional methods such as density gradient centrifugation. Clinical samples were then analyzed to validate the potential role of migrasome‐derived tumor biomarkers in lung adenocarcinoma. These findings underscore the promising applicability of our strategy for studying migrasomes in clinical disease diagnosis.

Recent advances in vision-language models (VLMs) have enabled broad progress in the general medical field. However, pathology still remains a more challenging sub-domain, with current pathology-specific VLMs exhibiting limitations in both diagnostic accuracy and reasoning plausibility. Such shortcomings are largely attributable to the nature of current pathology datasets, which are primarily composed of image–description pairs that lack the depth and structured diagnostic paradigms employed by real-world pathologists. In this study, we leverage pathology textbooks and real-world pathology experts to construct high-quality, reasoning-oriented datasets. Building on this, we introduce Patho-R1, a multimodal RL-based pathology Reasoner, trained through a three-stage pipeline: (1) continued pretraining on 3.5 million image-text pairs for knowledge infusion; (2) supervised fine-tuning on 500k high-quality Chain-of-Thought samples for reasoning incentivizing; (3) reinforcement learning using Group Relative Policy Optimization and Decoupled Clip and Dynamic sAmpling Policy Optimization strategies for multimodal reasoning quality refinement. To further assess the alignment quality of our dataset, we propose Patho-CLIP, trained on the same figure-caption corpus used for continued pretraining. Comprehensive experimental results demonstrate that both Patho-CLIP and Patho-R1 achieve robust performance across a wide range of pathology-related tasks, including zero-shot classification, cross-modal retrieval, Visual Question Answering, and Multiple Choice Question.

This study aimed to compare the diagnostic performance of 2 fat suppression techniques in diffusion-weighted imaging (DWI) for detecting and assessing focal liver lesions (FLLs): water excitation spectral heterogeneity adaptive radiofrequency pulses (WE-SHARP) and conventional spectral adiabatic inversion recovery (SPAIR). This prospective study enrolled eligible participants between October 2023 and August 2024. Various DWI techniques at 3T, SPAIR-DWI, WE-SHARP-DWI, and WE-SHARP-DWI with correction algorithms (WE-SHARP-DWI*), acquired at b values of 50, 400, 800, and 1200 s/mm², were used to evaluate FLLs. Two radiologists independently assessed several subjective image quality parameters: liver edge sharpness, vessel delineation, lesion conspicuity, fat suppression effectiveness, artifacts, and overall image quality. Signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and apparent diffusion coefficient (ADC) values were also measured. The diagnostic performance of all 3 sequences was evaluated using receiver operating characteristic (ROC) curve analysis. The study included 158 patients (67 with malignant and 91 with benign lesions) and 25 volunteers. Compared with SPAIR-DWI, the subjective image quality parameters were superior for both WE-SHARP sequences (P < .001). SNR increased 2.15-fold with WE-SHARP-DWI and 2.93-fold with WE-SHARP-DWI*. CNR also improved substantially with the WE-SHARP sequences (30.75 ± 36.83 and 42.43 ± 53.23 vs. 13.49 ± 14.39). Further, WE-SHARP sequences demonstrated lower ADC measurement variability with lower standard deviations (32.80 ± 19.13 × 10⁻³ mm²/s and 34.39 ± 18.22 × 10⁻³ mm²/s vs. 51.48 ± 17.89 × 10⁻³ mm²/s) in normal liver and more pronounced ADC differences between benign and malignant lesions (1179 × 10⁻3 mm²/s and 1197 × 10⁻3 mm²/s vs. 1009 × 10⁻3 mm²/s.). The WE-SHARP-DWI techniques demonstrated improved diagnostic performance with higher sensitivity (0.95 vs. 0.88) and greater area under the ROC curve (0.98 vs. 0.95) compared with SPAIR-DWI. Both WE-SHARP-DWI techniques demonstrated superior image quality and diagnostic value for assessing FLLs than the conventional SPAIR technique. These techniques retained clinically acceptable image quality even at high b values.

PURPOSE:To evaluate the test-retest repeatability of a rapid, free-breathing two-dimensional (2D) MR elastography (MRE) technique and to assess the reliability of the liver-stiffness measurements compared with conventional breath-hold MREs. METHODS:Fifteen and 115 participants were enrolled in the technical repeatability and measurement equivalence assessment cohorts, respectively. All participants underwent rapid free-breathing and conventional breath-hold 2D MRE (twice in repeatability cohort) on 1.5T scanners. Both methods have four phase offsets over one harmonic motion cycle at 60 Hz; one and 10 cycles were collected and processed in breath-hold and free-breathing MREs, respectively. The liver stiffness measurements of free-breathing (LSF) and breath-hold MRE (LSB) were calculated from manually drawn regions of interest. The repeatability coefficients and Spearman correlation were used to assess technical repeatability and measurement agreement between LSF and LSB. Univariable and multivariable linear regressions were performed to identify potential influencing factors, including age, sex, body mass index, and fat fraction, in the measurement agreement between LSB and LSF. RESULTS:The repeatability coefficient of free-breathing 2D MRE is comparable to breath-hold MRE (LSF: 20.8%, LSB: 20.4%). LSF showed a strong agreement and significant correlation with LSB (LSF = 1.01 × LSB, ρ = 0.94, p < 0.001). The measurement agreement between LSF and LSB was only significantly affected by sex (p = 0.047) after adjusting for confounding factors of age, body mass index, and fat fraction. CONCLUSIONS:The nongated, free-breathing, multislice 2D MRE technique can provide reliable liver stiffness measurements compared with conventional breath-hold 2D MRE. It could provide a comfortable alternative method with reliable liver stiffness measurements for patients who have difficulty in suspending respiration.

Migrasomes are newly discovered extracellular organelles released by migrating cells, such as immune cells, tumor cells, and other special functional cells like podocytes and embryonic cells. They contain a diverse array of constituents, including proteins, lipids, and RNA which can be released to the designated location to activate surrounding cells, thereby facilitating intercellular communication and signal transduction. Since then, our understanding of the mechanism and function of the migrasomes has expanded exponentially, with recent evidence indicating they are involved in various physiological and pathological processes, particularly in immune regulation. Furthermore, methods and techniques for extracting, detecting, and characterizing migrasomes are constantly advancing. Herein, we summarize the current understanding of migrasomes and their key roles in modulating immune responses, as well as the prospective challenges surrounding their clinical application, aiming to provide novel insights into the emerging organelles.

Accurate interpretation of human epidermal growth factor receptor 2 (HER2) immunohistochemistry (IHC) scores 0 and 1+ is crucial for treating HER2-low breast cancer patients with antibody-drug conjugates. To improve diagnostic precision, we developed models using 698 retrospectively collected HER2 IHC slides of breast cancer and tested them on an additional 501 slides reviewed by one junior and three senior pathologists. The artificial intelligence (AI)-based models included an invasive breast cancer (IBC) region segmentation model (Model I) and a nuclei detection model (Model II). Model I achieved mean intersection over union (MIoU) scores of 0.879 and 0.880 at 20× and 40× magnifications, and Model II's F1-scores were 0.866 and 0.878. The proposed AI microscope based on Models I and II achieved F1 scores of 0.878 and 0.906 and accuracies of 0.856 and 0.890 for interpreting IHC scores of 0 and 1+ at 20× and 40×, respectively, which was superior to that of a junior pathologist with an F1 score of 0.871 and an accuracy of 0.848. Additionally, the AI microscope showed high consistency with the interpretation results from the senior pathologists, reaching kappa values of 0.703 at 20× and 0.774 at 40×. This AI microscope has the potential to enhance the interpretation accuracy of HER2 IHC score in clinical settings.

Background: Clear cell renal cell carcinoma (ccRCC) is a prevalent urological malignancy, accounting for approximately 1.6% of all cancer-related deaths in 2022. While endocrine-disrupting chemicals (EDCs) have been implicated as risk factors for ccRCC, the toxicological profiles and immune mechanisms underlying Bisphenol A (BPA) exposure in ccRCC progression remain inadequately understood. Materials and Methods: Protein-protein interaction (PPI) analysis and visualization were performed on overlapping genes between ccRCC and BPA exposure. This was followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses to elucidate potential underlying mechanisms. Subsequently, 108 distinct machine learning algorithm combinations were evaluated to identify the optimal predictive model. An integrated CoxBoost and Ridge regression model was constructed to develop a prognostic signature, the performance of which was rigorously validated across two independent external datasets. Finally, molecular docking analyses were employed to investigate interactions between key genes and BPA. Results: A total of 114 overlapping targets associated with both ccRCC and BPA were identified. GO and KEGG analyses revealed enrichment in cancer-related pathways, including pathways in cancer, endocrine resistance, PD-L1 expression and PD-1 checkpoint signaling, T-cell receptor signaling, endocrine function, and immune responses. Machine learning algorithm selection identified the combined CoxBoost-Ridge approach as the optimal predictive model (achieving a training set concordance index (C-index) of 0.77). This model identified eight key genes (CHRM3, GABBR1, CCR4, KCNN4, PRKCE, CYP2C9, HPGD, FASN), which were the top-ranked by coefficient magnitude in the prognostic model. The prognostic signature demonstrated robust predictive performance in two independent external validation cohorts (C-index = 0.74 in cBioPortal; C-index = 0.81 in E-MTAB-1980). Furthermore, molecular docking analyses predicted strong binding affinities between BPA and these key targets (Vina scores all <-6.5 kcal/mol), suggesting a potential mechanism through which BPA may modulate their activity to promote renal carcinogenesis. Collectively, These findings suggested potential molecular mechanisms that may underpin BPA-induced ccRCC progression, generating hypotheses for future experimental validation. Conclusions: These findings enhance our understanding of the molecular mechanisms by which BPA induces ccRCC and highlight potential targets for therapeutic intervention, particularly in endocrine and immune-related pathways. This underscores the need for collaborative efforts to mitigate the impact of environmental toxins like BPA on public health.

Each year, millions of new cancer cases and cancer-related deaths underscore the urgent need for effective, affordable screening methods. Circulating tumor cells (CTCs), which derived from tumors and shedding into bloodstream, are considered promising biomarkers for liquid biopsy due to their unique biological significance and the substantial volume of supporting research. Among many advanced CTCs detection methods, electrochemical sensing is rapidly developing due to their high selectivity, high sensitivity, low cost, and rapid detection capability, well meeting the growing demand for non-invasive liquid biopsy. This review focuses on the entire procedure of detecting CTCs using electrochemical cytosensors, starting from sample preparation, detailing bio-recognition elements for capturing CTCs, highlighting design strategies of cytosensor, and discussing the prospects and challenges of electrochemical cytosensor applications.

Background:Improving the concordance of human epidermal growth factor receptor 2 (HER2) examinations among laboratories remains a challenge. In this multi-laboratory study, we investigated the concordance of HER2 immunohistochemistry (IHC) examination through manual and artificial intelligence (AI)-assisted interpretation. Methods:A tissue microarray (TMA) comprising 53 breast cancer samples was constructed and distributed to 35 participating laboratories. For each sample on every slide, IHC scores of 0, 1+, 2+, and 3+ were recorded. Subsequently, cases that failed to achieve complete agreement during manual interpretation were re-evaluated using an AI-assisted microscope. Results:During manual interpretation, 14 out of 53 cases (14/53, 26.4%) demonstrated concordant results across all laboratories, including 13 IHC-0 cases and 1 IHC-3+ case. Notably, cases scored as 1+ in at least one laboratory exhibited a low overall percentage agreement (OPA) and Fleiss Kappa value. Among the 39 cases with non-concordant manual interpretation, 14 cases (14/39, 35.9%) achieved complete agreement through AI-assisted HER2 interpretation. In cases where manual interpretation discrepancies were restricted to scores of 0 and 1+, 69.6% (16/23) of the cases still showed differences between 0 and 1+ in AI-assisted HER2 interpretation. Disagreements between manual and AI-assisted interpretation occurred significantly more frequently in sections manually scored as 1+ compared to those scored as 0 (58.6% vs. 2.1%, P<0.001). Conclusions:The weakly staining phenotype leads to poor agreement in the manual interpretation of HER2 IHC-1+ breast cancers. AI-assisted HER2 interpretation offers a viable approach for multi-laboratory studies, effectively avoiding the subjective errors inherent in manual interpretation.