Modern Pathology

Cervical cytology screening relies on the visual assessment of nuclear and cellular morphology across thousands of cells per slide, a labor-intensive task that could benefit from quantitative computational support. Here we present CytoLens, a modular deep learning system that detects and counts nuclei, delineates nuclei and whole cells, measures the nuclear-to-cell area ratio (N:C), classifies cells from multifocal images and scores candidate abnormality. We trained and tested each module on a public dataset matched to its task, without pooling datasets. CytoLens detected nuclei with an F1 score of 0.931 (95% confidence interval (CI), 0.914-0.946) and counted 75 of 100 fields to within one nucleus. Point-prompted segmentation achieved mean Dice coefficients of 0.901 for nuclei and 0.847 for whole cells and, without retraining, transferred to a second dataset in which measured N:C increased with lesion grade in parallel with expert outlines. Multifocal hierarchical classification reached a mean accuracy of 0.947 across four Bethesda-aligned categories, and abnormality scores ranked annotated abnormal candidates with an area under the receiver operating characteristic curve (AUROC) of 0.865. These results support the evaluation of CytoLens on patient-linked clinical specimens.

Background: General-purpose large language models (LLMs) are increasingly evaluated in diagnostic pathology, but prior studies have largely emphasized diagnostic accuracy rather than how models fail. We evaluated four LLMs for diagnostic performance, pathology-relevant errors and hallucinations, their burden, and potential clinical impact across multi-organ pathology cases. Design: In this retrospective multicenter study, 153 pathology cases from two institutions spanning 20 organs were evaluated using ChatGPT-5.3 (LLM1), Gemini 3 (LLM2), Grok 4.20 (LLM3), and Claude Opus 4.6 (LLM4). Each LLM received multi-magnification histologic images with clinical context and generated a microscopic description and diagnosis. No data splitting, training, or fine-tuning was performed. Twenty-one pathologists assessed 612 outputs for diagnostic correctness, error and hallucination type and burden, clinical impact (0-4), and overall performance (1-5). Results: Strict diagnostic accuracy was 48.9% overall (60.1% including partially correct diagnoses) and ranged from 36.6% to 58.2% across LLMs. Errors occurred in 82.0% of outputs and hallucinations in 76.6%; 90.2% contained at least one error or hallucination. Misinterpretation was the most frequent error (72.1%), while fabricated histologic features were the dominant hallucination type (74.5%). LLM2 had significantly lower misinterpretation rates than the other three LLMs, while LLM3 had significantly higher fabricated-feature hallucination rates than LLM1 and LLM2. Errors were more frequent in incorrect than correct diagnoses (97.5% vs 66.2%), as were hallucinations (95.1% vs 59.9%; both p < 0.001). Every strictly incorrect diagnosis contained an error and/or hallucination, while 79.9% of strictly correct diagnoses also contained at least one. In multivariable analysis, LLM3 was independently associated with higher odds of an incorrect diagnosis. Diagnostic correctness was the strongest determinant of high clinical impact; each one-point increase in hallucination burden increased the odds by 63% (adjusted OR, 1.63). Conclusion: Diagnostic accuracy alone substantially underestimates the safety limitations of general-purpose LLMs in histopathology. Errors and hallucinations were common even when the final diagnosis was correct, and their burden was independently associated with clinical impact. Evaluation frameworks should therefore assess both diagnostic correctness and the reliability and potential consequences of accompanying generated content.

Current state of the art computational pathology foundation models attain satisfying accuracy most general pathology tasks including detection of pancreatic ductal adenocarcinoma (PDAC) but fail to provide sufficient explainability, operating as "black boxes". Some interpretability effort has been made, with models providing attention maps that project back on the slide the models high attended areas that pushed for the given prediction result, without further naming or reasoning explanation which can limit pathologists trust toward the prediction. We developed C3PRO, a cell-centric framework attempting to provide deeper understanding and explainability, by using embeddings extracted using a pathology foundation model, from cell centered tiles containing both the cell and its close neighborhood, which were then clustered into recurrent morphological phenotypes. A spatial graph was additionally constructed on top of the back-projected vocabulary to account for recurrent motifs. The evaluation of the model performance involved a cohort of 809 patients (825 diagnostic biopsies) using nested cross validation. C3PRO was benchmarked against four different multiple instance learning frameworks and achieved on-par performance for diagnostic (0.958 AUC), survival (0.617) and resectability (0.696) prediction, while providing additional insights compared to the strongest baseline models. Each clustered phenotype was probed and extracted as a representative tile mosaic and annotated by two expert pathologists, constituting a phenotype vocabulary available for statistical analysis. These phenotypes were scored by pathologists with physical brightfield microscopy and replicated the independent statistical significance impact on overall survival prediction from diagnostic biopsies.

Background and Objective: Micro-ultrasound (MicroUS) is an emerging high-resolution imaging modality for prostate cancer diagnosis and real-time biopsy guidance, but its automated analysis is limited by labor-intensive expert annotation. We aimed to develop and evaluate a foundation model for data-efficient MicroUS image analysis. Methods: We developed muFM, a MicroUS-specific foundation model pretrained using DINOv2-style self-supervised learning on 1,736,391 unlabeled images from 780 patients. The model was fine-tuned and evaluated in three downstream cohorts for prostate capsule segmentation (public, 75 patients), urethra segmentation (institutional, 105 patients), and prostate cancer lesion localization (institutional, 190 patients). Under limited annotation, muFM was compared with randomly initialized models and foundation models pretrained on natural, medical, and ultrasound images. Evaluation used the Dice scores and 95th-percentile Hausdorff distance (HD95) for segmentation and free-response receiver operating characteristic analysis for lesion localization. Key Findings and Limitations: muFM improved performance across all three tasks, with the largest gains under limited annotation. With 10% of training scans, muFM achieved Dice scores of 93.8% for prostate capsule segmentation and 58.4% for urethra segmentation, versus 87.9% and 47.0% for the strongest comparators. Corresponding HD95 values were 2.0 and 4.3 mm. For lesion localization, muFM outperformed other models with approximately 0.75 sensitivity at two false-positive detections per 3D volume. Limitations include retrospective single-institution pretraining and lack of explicit 3D modeling. Conclusions and Clinical Implications: MicroUS-specific self-supervised pretraining improved data-efficient image analysis across prostate and urethra segmentation, and lesion localization. This approach may reduce annotation requirements for developing automated tools in MicroUS image analysis.

Foundation models (FMs) have rapidly become dominant in artificial intelligence and are increasingly being adopted in computational pathology. Numerous pathology-specific FMs have been developed and evaluated for a variety of downstream tasks, most commonly using frozen image embeddings with linear probes. While a pathology FM, at least implicitly suggests broad reusability across tasks, the transferability of these models to neurodegenerative disease-related tasks remains largely unexplored. In this study, we evaluated fourteen frozen feature extractors, including general-vision and pathology FMs, across four pathology image classification datasets spanning two specialized neurodegenerative disease immuno-histochemistry (IHC) tasks (Tau neurofibrillary tangle (NFT) and amyloid-{beta} plaque classification) and two cancer-related hematoxylin and eosin (H&E) tasks (the breast tissue BACH dataset and the multi-tissue TIL dataset). The cancer-related H&E datasets represent tasks more closely aligned with the predominant pretraining domain of many current pathology FMs, whereas the neurodegenerative IHC datasets represent a more specialized domain with limited representation in existing FM pretraining cohorts. Frozen embedding linear probes were compared against a conventional supervised ResNet-50 convolutional neural network (CNN) trained directly on image tiles using identical train, validation, and test splits. Across both neurodegenerative IHC datasets, the supervised CNN substantially outperformed all frozen FM linear probes. In contrast, pathology FMs achieved performance comparable to, and in some cases exceeding, the supervised CNN across the two cancer-related H&E datasets, with UNI2-h achieving the highest performance on BACH and several pathology FMs performing on par with the CNN on the larger TIL dataset. Furthermore, a supervised CNN trained using only 1% of the Tau NFT training data (1,938 tiles) still exceeded the performance of the best frozen FM linear probe trained on the complete dataset. Together, these results suggest the transferability of frozen pathology FMs may depend strongly on how well the downstream task is represented by their pretraining domain. These findings demonstrate frozen pathology FMs transfer effectively to the cancer-related H&E tasks evaluated here but may be less effective for specialized neurodegenerative IHC tasks less well represented in current FM pretraining cohorts. Expanding pathology FM pretraining datasets to include a broader range of disease domains and staining modalities may therefore improve transferability to specialized pathology applications. Our findings also highlight the continued importance of conventional supervised learning and motivate future work investigating nonlinear probes and end-to-end FM fine-tuning.

Background: Clinical registries and institutional quality control (QC) in pathology heavily depend on complex histopathological data. However, critical tumor characteristics are mostly locked in unstructured free-text reports. We hypothesize that locally deployed, resource-efficient language models operating within a secure institutional network can reliably extract relevant variables from German reports of hepatocellular carcinoma (HCC), thereby enabling automated data curation without transferring sensitive patient data. Methods: We developed a local, privacy-preserving open-weight Large Language Model (LLM) pipeline consisting of two consecutive stages: Stage 1 utilizes a lightweight gatekeeper model to isolate the target cohort of confirmed Hepatocellular Carcinoma cases from a screening pool of N = 2,864 reports. Stage 2 deploys a high-capacity model to extract 15 predefined histopathological variables, including tumor-node-metastasis staging, histological grading, vascular invasion, and selected immunohistochemical features into structured JSON format. For both Stages, pipeline performance was evaluated using Accuracy, Macro F1-scores, and Cohen's kappa compared to expert-curated reference annotations. Results: The top-performing Stage 1 gatekeeper isolated the target patient cohort with zero classification errors, effectively reaching complete agreement with the reference standard (ground truth) and correcting inaccurate administrative metadata. For Stage 2 feature extraction, current-generation models demonstrated robust structural reliability, reducing the burden of manual post-processing. The 31B model configuration (Gemma 4) virtually eliminated objective factual errors, achieving near-perfect agreement with the reference standard (Macro F1 > 0.99). Conclusion: Local two-stage pipelines offer a highly accurate, resource-efficient, and secure solution for clinical text processing. This framework serves as a privacy-preserving engine for institutional quality control and the scalable curation of pathology registries.

Purpose: To evaluate whether visually assessed high-quality prostate MRI, as defined by the Prostate Imaging Quality (PI-QUAL) scoring system, is sufficient to ensure consistent AI model performance, or whether hidden acquisition-related technical effects introduce variability that cannot be detected through visual inspection. Methods: This retrospective study included 568 bpMRI scans from seven institutions, classified using PI-QUAL v2 into high-quality (scores 2-3) and low-quality (score 1) groups. Quantitative image quality features were extracted using MRQy from T2w and ADC sequences. Three experiments were conducted on high-quality scans: (1) unsupervised clustering to identify hidden technical batch effects; (2) V-Net models (M) for prostate capsule segmentation on T2w using single-cluster and balanced multi-cluster training strategies; and (3) VoxelMorph-based deformable registration (M2) to align ADC maps to T2w images. Performance metrics included DSC, precision, recall, volume ratio, HD95, ASSD, and TRE. Kruskal-Wallis tests assessed cross-cluster variability, Two One-Sided Tests (TOST) evaluated performance equivalence, and bootstrap resampling quantified training efficiency. Results: Clustering identified four clusters among 277 high-quality bpMRI scans. Single-cluster models showed variable performance (DSC=0.41-0.85) with significant batch effects (p<0.001), even with 140 training samples. The balanced multi-cluster model (n=60) eliminated batch effects (p=0.075), achieved equivalent performance (p<0.001), and demonstrated 2.23-fold greater training efficiency despite using 2.3-fold fewer samples. Registration performance also varied significantly across clusters (TRE=0.86-5.47 mm, p<0.001). Conclusion: Acquisition-related batch effects significantly impact AI generalizability in prostate MRI, even among high-quality scans. Quantitative clustering enables detection of technical variations and supports diversity-aware training strategies that achieve equivalent performance with 2.3-fold fewer samples while eliminating batch effects.

Epidermodysplasia verruciformis (EV) is characterized by susceptibility to cutaneous {beta}-human papillomaviruses, persistent viral lesions and an increased risk of cutaneous squamous cell carcinoma (cSCC). We investigated whether expression of Toll-like receptors 3, 4, 5 and 9 (TLR3, TLR4, TLR5 and TLR9) differs across histological states within EV. Immunoreactivity was quantified by digital H-score using QuPath in 97 specimens from 28 patients, comprising 13 lesion-free skin samples, 15 flat warts and 69 cSCCs. Prespecified comparisons evaluated flat wart versus lesion-free skin, cSCC versus non-neoplastic tissue and invasive versus in situ cSCC, with invasive tumour differentiation assessed secondarily. Patient-clustered generalized estimating equations and false-discovery-rate (FDR) correction were used for inference. After correction for multiple testing, TLR9 H-scores were higher in flat warts than in lesion-free skin ({beta}=60.58; q=0.0332). Compared with non-neoplastic tissue, cSCC showed lower TLR3 ({beta}=-81.40; q=0.00745), higher TLR4 ({beta}=55.59; q=0.04793) and lower TLR9 expression ({beta}=-34.80; q=0.03412). TLR5 showed a nominal increase in cSCC that did not meet the FDR threshold. None of the four receptors differed significantly between invasive and in situ cSCC, and no differentiation trend remained significant after FDR correction. These findings show receptor-specific differences in TLR expression across histological states of EV, particularly involving TLR3, TLR4 and TLR9, although the diagnostic and prognostic significance of these tissue patterns remains to be established.

Predicting molecular receptor status, including estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), directly from routine hematoxylin and eosin (H&E)-stained whole-slide images (WSIs) could reduce reliance on costly immunohistochemistry, but image-only models have no access to the transcriptomic programs that define these subtypes. Here we show that coupling a whole-slide vision transformer (GigaPath) with bulk RNA-sequencing profiles through a training-time-only contrastive objective improves receptor-status classification from WSIs alone, without requiring any RNA-seq data at inference. Rather than training a dedicated bulk-transcriptomics encoder, we purpose the gene-sentence encoding and pretrained text encoder from BioBERT1 and apply them to bulk RNA-seq profiles from The Cancer Genome Atlas breast cancer cohort (TCGA-BRCA). Under five-fold cross-validation, this RNA-guided pre-training consistently improved AUROC for ER, PR, and HER2 relative to a contrastive-pre-training-free baseline and reproducibly introduced receptor-relevant structure into the learned slide representation. However, evaluation on independent external cohorts revealed a striking, receptor-dependent divergence: gains for ER and PR, both governed by broad, multi-gene luminal transcriptional programs, generalized robustly, whereas the apparent AUROC improvement for HER2, driven by focal ERBB2 amplification rather than a coordinated transcriptional signature, masked a severe loss of classification sensitivity at the operating threshold. These findings establish training-time transcriptomic supervision as a practical route to enhancing histology-only molecular subtyping, while showing that gene-selection strategy, and not model architecture alone, determines which molecular alterations such supervision can teach a vision model to recognize.

Purpose To assess whether follow-up MRI, biopsy status, and timing of cribriform disease detection predict active surveillance (AS) failure in a contemporary transatlantic cohort. Materials and Methods. Retrospective MRI-driven AS cohort of 619 men at two academic centers (UK/USA), 2012-2024. Primary outcome: AS failure (disease-specific mortality, metastasis, grade group [GG] [&ge;]4 progression, or post-treatment biochemical recurrence). Key exposures: final MRI and biopsy status, timing of cribriform [&ge;]GG2 detection. Cumulative incidence was derived; independent predictors used multivariable Fine-Gray models. Results. Over median follow-up of 6.4 years, 39/619 men (6%) experienced AS failure. Cribriform [&ge;]GG2 was detected in 82/619 (13%) on follow-up biopsy versus 24/619 (4%) at baseline, comparably in baseline GG1 (10%) and non-cribriform GG2 (9%), and only in MRI-visible tumors. Three-year AS failure after final biopsy was 31% in men with cribriform [&ge;]GG2 versus 0-2% with stable or non-cribriform disease; 44% of failures had stable MRI. On Fine-Gray analysis, cribriform [&ge;]GG2 was the only independent predictor of AS failure in re-biopsied men (SHR 30; 95% CI 5.2-167; P<0.001); no type of MRI progression category was independently predictive. Among 595 men cribriform-negative at baseline, 6-year AS failure after a second biopsy was 34% if cribriform [&ge;]GG2 was detected versus 7% if still cribriform-negative. Conclusions. Cribriform status at follow-up biopsy independently predicts AS failure, whereas MRI progression does not. The three-fold higher cribriform prevalence at follow-up likely reflects baseline sampling error. Because newly detected cribriform tumours were exclusively MRI-visible, early repeat biopsy of patients with MRI-visible disease may considerably improve risk stratification.

Patients with majority variant subtype histology including sarcomatoid urothelial carcinoma (SUC) have poor outcomes and limited treatment options, and there are few preclinical resources to investigate rare histologic subtypes. We developed three novel variant subtype histology bladder cancer (BC) cell lines by finding that basal BC tumor cells derived from mice treated with N-butyl-N-(4-hydroxybutyl)-nitrosamine (BBN) were initially not competent to grow in immune competent mice, but then acquired tumorigenicity following passaging (recycling) in mice lacking mature B and T cells. Three genitourinary oncology pathologists identified BBN966 as SUC, BBN964 as a hybrid between SUC and squamous, and BBN975 as squamous, and each had unique growth characteristics. Using a publicly available dataset and our spatial transcriptomics analysis of a cohort of human tumors with mixed SUC and UC as comparators, we found that consistent with human SUC, BBN966 lack expression of basal and luminal genes. Consistent with human SUC and spatial transcriptomics analysis of canine SUC, BBN966 had the highest epithelial to mesenchymal transition (EMT) upregulation, and all three recycled tumor cells had increased EMT pathway expression in recycled tumor cells. Moreover, increased expression of genes upregulated in all three recycled tumor cells associated with worse overall survival in patients with muscle invasive UC. Similar to human SUC, BBN966 tumors had higher proportion of M2 macrophage gene expression. Finally, we found in the largest cohort of PD-L1 IHC to date that 20 of 38 (52.6%) human SUC had CPS > 50, which was driven mostly by tumor proportion score. Consistent with this, we found anti-PD1 immune checkpoint inhibitor (ICI) significantly inhibited BBN966 tumor growth, while the hybrid BBN964 tumors had mixed response to ICI. Thus, the murine BBN966 tumors exhibited similarities to human SUC, and can be used to examine novel therapies for translation into future SUC clinical studies.

Chimeric antigen receptor (CAR) T cells are patient-derived T cells engineered to recognize specific antigens on malignant cells, representing a major breakthrough in personalized cancer immunotherapy. While CAR-T cell therapy is an established treatment for relapsed or refractory hematologic malignancies, understanding the rationale and methodology behind its development is essential for training the next generation of oncology researchers. Here, we describe a novel, hands-on small-enrollment laboratory course designed to teach upper-level undergraduate and graduate students about immunotherapies and CAR-T cell technology by paralleling a typical bench-scale development process. Students isolated T cells from porcine peripheral blood mononuclear cells (PBMCs) via immunomagnetic selection, followed by flow cytometry to assess purity, viability, and yield. Students then used pre-engineered B7H3 CAR-T cells to perform co-culture assays with U87 glioblastoma cells to evaluate tumor cell killing efficacy. Cytotoxicity was qualitatively evaluated using immunofluorescence microscopy, while IFN-gamma release was quantified using Enzyme-linked Immunosorbent Assay (ELISA). Student performance was evaluated through electronic laboratory notebooks, discussion questions, a capstone project, and student-generated data. Our results demonstrate students high academic performance across assignments and the successful execution of complex laboratory workflows. Collectively, these findings establish the feasibility and implementation of a research-integrated curriculum providing foundational exposure to immuno-oncology research tools and CAR-T cell production techniques.

Background: Epigenetic deregulation is a defining feature of leukemia, and DNA methylation and chromatin accessibility profiles encode the cell of origin, lineage and clinical behaviour of the disease. Artificial intelligence (AI) and machine learning (ML) can extract predictive patterns from these high-dimensional data, but the published models have not been systematically appraised. We systematically reviewed AI models built on epigenomic data in leukemia for diagnosis, prognosis, relapse and treatment-response prediction, and measurable residual disease (MRD) detection. Methods: This review followed PRISMA 2020 and the SWiM guideline. PubMed was searched on 7 October 2026 with three structured queries, supplemented by a broader PubMed query, citation searching and web searching. Eligible studies were peer-reviewed primary studies that applied an ML or deep-learning algorithm to DNA methylation, chromatin-accessibility or histone-modification data from human leukemia samples and reported a quantitative diagnostic, prognostic, response or MRD outcome. Risk of bias and applicability were assessed with PROBAST, informed by PROBAST+AI. Owing to heterogeneity, results were synthesized narratively and in structured tables, without meta-analysis. Results: Of 68 unique database records and 32 reports from other sources, 30 studies (2010-2026; 22 since 2020) were included: 16 in acute myeloid leukemia, 4 in acute lymphoblastic leukemia, 2 spanning acute leukemia lineages, 3 in juvenile myelomonocytic leukemia and 5 in chronic lymphocytic leukemia; 10 were pediatric. DNA methylation was the input in 28 studies and chromatin accessibility in 2; no eligible study modelled histone marks. Tree ensembles (n = 9) and support vector machines (n = 5) were the most frequent algorithms. Diagnostic classifiers reported accuracy, concordance or F1 scores of 0.87-1.00, and two classifiers were applied to nanopore sequencing data, one returning acute leukemia subtype calls within about 2 hours. Prognostic epigenomic scores were associated with survival (reported hazard ratios 2.36-10.82), but discrimination fell on transport to independent cohorts (C-index 0.677 internally vs 0.529 in one external cohort). Fifteen studies validated models on an external epigenomic cohort, one prospectively. Overall risk of bias was high in 21 studies, unclear in 8 and low in 1, driven by the analysis domain (small samples, absent calibration, possible data leakage). Calibration was reported in 3 studies and code was available for 11. Conclusions: Epigenomic AI models classify leukemia subtypes with high reported accuracy and can deliver same-day diagnoses, and methylation-based scores add prognostic information. However, the evidence is dominated by retrospective, high-risk-of-bias development studies with limited external validation and almost no prospective or impact evaluation. Prospective, multi-center validation with calibrated, transparently reported models is required before clinical adoption.

Diffusion-weighted images are sometimes unavailable or degraded in prostate MRI. We asked whether T2-weighted MRI alone retains usable cancer-related signal and whether the computation can be audited. An image-level transfer-learning ensemble reached out-of-fold AUC 0.883 (95% CI 0.836-0.924) on 243 de-identified institutional T2 images; a handcrafted image-level model reached 0.924 (0.889-0.953) and, in an exploratory external test of 137 Prostate158 patients, 0.643 (0.545-0.734) against 0.494 previously. A volume-level DINOv2-attention model reached AUC 0.717 (0.572-0.853) in 133 held-out biopsy-linked patients, 0.733 with test-time augmentation, and a locked precursor 0.618 (0.566-0.668) in 561 independent PROMIS patients. Distillation reproduced the frozen teacher with R-squared 0.776, a transparent boosting machine 0.678 and a readable model tree 0.550, exposing dependence on gland size, low T2 signal and texture. A bounded dual-agent application gates input quality and anatomy before frozen inference, returning audit metadata. These findings support further evaluation of auditable T2-only analysis when complementary sequences are unavailable or degraded.

High-risk HPV is central to cervical carcinogenesis, but whether individual genotypes drive distinct tumor biology or converge on a shared oncogenic state remains unclear. Profiling viral and host gene expression across two independent cohorts (n = 93; 76 HPV16, 17 HPV18), we show that HPV16- and HPV18-associated cervical cancers differ in host transcriptional programs. A 55-gene signature discriminated the two genotypes, and 293 genes were concordantly differentially expressed across cohorts. Pathway and transcription-factor activity inference organized these differences along genotype-specific axes: HPV16 tumors were enriched for epithelial polarity and RTK-RAS-MAPK signatures, with elevated inferred CDX2 regulon activity, a transcription factor not normally active in cervical epithelium, suggesting an aberrant lineage-identity program not previously linked to this genotype. HPV18 tumors instead showed metabolic reprogramming and inflammatory signaling (STAT3, HIF1A), alongside reduced CASP8 and increased FDX1 expression, a pattern consistent with sensitization to immunogenic cell death. Both genotypes also organized along a continuum of early-to-late viral gene expression correlating with distinct host pathways: extracellular-matrix and translation-related in HPV16, interferon/antiviral and cell-cycle/mitochondrial in HPV18, indicating that viral transcriptional state associates with host biology beyond genotype alone. These results point to candidate genotype-specific vulnerabilities that merit functional validation.

Automated triage of whole-slide cervical cytology images remains constrained by the heavy dependence on single-cell bounding boxes or pixel-level annotations, which introduce significant bottlenecks, high expert overhead, and observer variability. To address these challenges, we present a weakly-supervised, detection-free Multiple Instance Learning (MIL) framework powered by a Gated Attention pooling engine that operates exclusively on slide-level predictions. By treating whole-slide images as bags of local instance patches, the dual-branch gated network dynamically assigns non-linear attention weights to highlight isolated dysplastic cells while suppressing benign background and debris, successfully resolving the "needle-in-a-haystack" problem inherent to high-grade lesions. Evaluated across multi-center cohorts including internal datasets (SIPaKMeD, Herlev, and CRIC) and the unannotated, out-of-distribution Mendeley LBC validation cohort processed via an unsupervised marker-controlled watershed pipeline our approach demonstrates robust generalization without requiring dense instance labels. Furthermore, our architectural comparison reveals a key trade-off: lightweight backbones like MobileNetV2 optimize in-distribution multi-center accuracy (90.96% Accuracy, 0.9800 ROC-AUC), whereas higher-capacity models like Xception ensure superior out-of-distribution robustness (80.43% Balanced Accuracy) under domain shifts. Ultimately, this detection-less pipeline provides a scalable and clinically viable solution to automate slide triage and alleviate cytopathologist shortages in resource-constrained environments.

Background: Archival biobanks of formalin-fixed paraffin-embedded (FFPE) specimens represent a vast and underused resource for retrospective molecular studies linked to clinical follow-up. However, fixation-induced chemical modification and nucleic acid fragmentation limit scalable genomic and transcriptomic profiling. FFPE-CUTAC (Cleavage Under Targeted Accessible Chromatin) is an RNA polymerase II-targeted DNA assay specifically designed to profile regulatory activity in FFPE tissue. Here, we assess its archival robustness, information content, compatibility with RNA-seq, and ability to recover copy-number alterations. Results: In meningioma and breast cancer cohorts, we benchmarked FFPE-CUTAC against matched fresh-frozen RNA-seq, FFPE RNA-seq and FFPE whole-genome sequencing. For FFPE specimens spanning up to 25 years, FFPE-CUTAC showed no systematic collection-year-associated decline in library yield or signal-quality metrics. In contrast, FFPE RNA-seq is more affected by the FFPE tissue degradation and specimen age on read composition, gene-body coverage, and gene detection. By measuring RNA polymerase II occupancy on chromatin rather than mature RNA abundance, FFPE-CUTAC is less constrained by transcript half-life, polyadenylation or transcript-capture design. FFPE-CUTAC extends beyond gene regions to non-coding regulatory elements, while the shared gene-level signals supported its integration with existing RNA-seq cohorts for large-scale, long-term clinical association study. FFPE-CUTAC libraries also retain DNA dosage information and can be used to recover chromosome-arm gain, loss and intact states with 96% concordance to matched whole-genome sequencing. Conclusions: By converting routinely preserved pathology sections into integrated regulatory and copy-number profiles, FFPE-CUTAC provides a practical foundation for constructing clinically annotated disease maps to support molecular stratification, prognostic modeling and treatment-association studies.

Purpose: A minority of patients with bladder cancer derive lasting benefit from immunotherapy, and current biomarkers capture only part of the biology. We evaluated whether tumor-cell MHC class II defines an immune state associated with outcome across disease stages and after PD-1/PD-L1 blockade. Experimental Design: We localized differential MHC-II activity to malignant epithelial cells by single-nucleus and single-cell RNA sequencing and confirmed HLA-DR expression by immunohistochemistry in 122 specimens. Chromatin profiling and IFN-{gamma} stimulation of cell lines and primary tumors supported derivation of an 11-gene tumor-cell program. Associations were tested in 434 patients with non-muscle-invasive disease, 82 receiving neoadjuvant pembrolizumab, and 288 receiving atezolizumab for metastatic disease. Results: Tumor-cell MHC-II was present in approximately one third of bladder cancers, and malignant cells accounted for the differential signal between MHC-II-high and MHC-II-low tumors. The program was inducible by IFN-{gamma} through JAK/STAT signaling. Within luminal non-muscle-invasive disease, MHC-II-low status was associated with progression (HR 3.31, 95% CI 1.37 to 7.99). In PURE-01, the 11-gene program was associated with pathological complete response (52% versus 24%, p = 0.018) and recurrence-free survival (p = 0.0054), retaining an independent association in models including tumor mutational burden and PD-L1. In bladder-primary metastatic disease, program-high status was associated with overall survival in unadjusted analysis (HR 0.62, 95% CI 0.44 to 0.89); the association did not extend to upper tract tumors (interaction p = 0.0018). Conclusions: Tumor-cell MHC-II defines an inducible state associated with outcomes across bladder cancer stages and after checkpoint blockade, supporting prospective evaluation by RNA and immunohistochemistry.

Aim: Brucellosis remains endemic throughout the Mediterranean basin, the Middle East and Central Asia. Genitourinary involvement is an uncommon but clinically consequential focal complication that closely mimics urological emergencies and may precipitate unnecessary surgery. Contemporary data describing its frequency, laboratory signature and treatment outcomes in endemic settings remain limited. Material-Methods: We retrospectively screened all adults ([&ge;]18 years) with confirmed brucellosis managed at a tertiary infectious diseases unit between January 2008 and December 2020. Genitourinary involvement was defined as epididymo-orchitis, orchitis, epididymitis and/or brucellar pyelonephritis. Clinical, laboratory, ultrasonographic, therapeutic and outcome variables were abstracted using a standardised form. Proportions are reported with Wilson score 95% confidence intervals (CI). Reporting followed the STROBE statement. Result: Genitourinary involvement was identified in 18 of 125 patients with brucellosis (14.4%; 95% CI 9.3-21.6); 16 (88.9%) were men and the mean age was 35.2 years (range 22-62). Scrotal disease dominated the male presentation: epididymo-orchitis in 10 of 16 men (62.5%; 95% CI 38.6-81.5), orchitis in 4 (25.0%) and epididymitis in 2 (12.5%). Both women presented with pyelonephritis, one of whom developed a tubo-ovarian abscess requiring drainage. Constitutional features were near-universal (fatigue 100%, night sweats 88.9%, fever 77.8%). Leukocyte counts remained within the reference interval in every patient (mean 5,680/mm3), whereas erythrocyte sedimentation rate (mean 32 mm/h) and C-reactive protein (mean 24.2 mg/L) were only mildly to moderately raised. The Coombs anti-Brucella test was positive in 18 of 18 patients (100%) and the standard agglutination test in 17 of 18 (94.4%), whereas blood culture yielded Brucella melitensis in only 1 of 18 (5.6%). Concomitant extragenitourinary focal disease was present in 8 patients. All patients received doxycycline plus rifampicin; none required modification of the regimen or orchiectomy, and no treatment failure occurred (0 of 18; 95% CI 0-17.6). Median follow-up was 12 weeks (range 6-52) and no relapse was documented within this interval. Conclusions: In endemic regions, genitourinary brucellosis should be suspected in any patient with acute scrotal or flank symptoms accompanied by constitutional complaints, even when the leukocyte count is normal and acute-phase reactants are only marginally elevated. Coombs anti-Brucella serology rather than blood culture is the diagnostic cornerstone, and prompt doxycycline-rifampicin therapy averts surgical intervention.

Prostate cancer (PC) remains a significant health concern as the second leading cause of cancer-related death in American men. Androgen deprivation therapies represent the standard of care for patients with PC, regardless of disease stage. Gonadotropin-releasing hormone (GnRH) antagonists, such as Degarelix, are commonly used hormone therapies for PC and work by suppressing testosterone production. Previous studies indicate a prostate involution-regeneration cycle following surgical castration and testosterone addback. However, whether a similar response occurs following pharmacologic androgen deprivation with agents such as Degarelix remains poorly understood. In particular, the temporal dynamics of prostate involution and subsequent regeneration after chemical castration have not been systematically defined. Our study utilizes immunocompetent mouse models to evaluate the systemic and temporal effects of Degarelix treatment and determine its efficacy in suppressing PC tumor growth. We show that continuous Degarelix treatment significantly suppressed systemic testosterone levels in mice, whereas testosterone levels returned to physiological baseline following Degarelix withdrawal. Gross anatomical, H&E, and Trichrome analyses demonstrated that Degarelix induced castration resulted in prostate involution, characterized by reduced gland size and increased stromal remodeling. This phenotype was reversed after Degarelix withdrawal, highlighting the reversibility of pharmacological androgen deprivation. Immunofluorescence staining revealed changes in basal and luminal epithelial cell populations within the prostatic gland, indicating changes in prostate glandular architecture following Degarelix induced castration and subsequent regeneration. Next, we evaluated the impact of chemical castration on PC tumor growth. Although Degarelix initially suppressed tumor growth, this response was not sustained through the study endpoint, suggesting the emergence of acquired resistance to androgen deprivation. These findings establish this immunocompetent model as a physiologically relevant platform for dissecting the mechanisms that drive therapeutic resistance and tumor progression following androgen deprivation. Our findings highlight the effectiveness of chemical castration with Degarelix and establish a robust model to study both therapeutic responses and the emergence of castration resistance in PC.