Artificial intelligence (AI) is becoming a clinical tool for prostate pathology, but generalization across variations in sample preparation and preservation over prolonged time periods remains poorly understood. We evaluated GleasonAI, an end-to-end attention-based multiple instance learning model, on an independent validation cohort comprising 10,366 biopsy cores from 1,028 patients across 14 Swedish regions, using archival diagnostic specimens from the ProMort cohorts collected between 1998-2015. The model achieved an overall quadratic-weighted kappa of 0.86 for core-level ISUP grading, comparable to several experienced pathologists and consistent across geographic regions. Notably, performance remained stable across the 17-year collection period, demonstrating robustness to time-related variation in archival material, a property not consistently observed with foundation model-based approaches, with exploratory analysis demonstrating a significant prognostic gradient across AI-assigned grade groups for prostate cancer-specific mortality. These findings support the generalizability of the AI grading model and demonstrate the potential of pathology archives as a large-scale resource for AI development, validation, and retrospective prognostic research.

Introduction:Diagnostic work-up of peripheral pulmonary lesions (PPLs) remains a challenge in interventional pulmonology. Conventional bronchoscopy and trans-thoracic needle aspiration (TTNA) often entail limitations in accuracy or safety. Robotic-assisted bronchoscopy with shape-sensing technology (ssRAB) combined with artificial intelligence (AI)-aided augmented fluoroscopy (LungVision system) represents an innovative approach to enhance lesion localization and diagnostic yield for small or otherwise hard-to-reach lesions while maintaining the safety profile of the procedure. Materials and methods:This retrospective observational study included all procedures performed under general anesthesia using the ssRAB (Ion™ robotic platform) in combination with C-arm based computed tomography (LungVision System) at the Interventional Pulmonology Unit of IRCCS Azienda Ospedaliero-Universitaria, Policlinico Sant'Orsola, Bologna (Italy), between December 2024 and September 2025. Overall, 82 patients with 94 sampled pulmonary lesions were included; in 12 patients, two distinct nodules were sampled during the same bronchoscopic procedure. Lesions features, procedural characteristics, diagnostic yield and complications were collected and statistically analyzed. Diagnostic yield was defined according to the 2024 Delphi consensus and STARD 2015 guidelines. Results:This study included 82 patients and 94 sampled pulmonary nodules. The median size of the lesion was 14 mm (IQR 11-18 mm). Target lesions were identified by radial endobronchial ultrasound (r-EBUS) in 80% of cases. A definitive cyto-histologic diagnosis was achieved in 74.5% of cases. Lesions ≥10 mm yielded a 79.2% diagnostic rate, versus 59.1% for smaller nodules. Diagnostic success was independently predicted by upper lobe location (OR 3.28; p = 0.031) and lesion visibility on CABT (OR 3.22; p = 0.043). The overall complication rate was 3.3% (three pneumothorax cases; no major bleeding). The median procedure time was 34.5 min, and most procedures were performed in a day-hospital setting. Discussion and conclusions:The integration of robotic-assisted bronchoscopy with the LungVision system exhibited high diagnostic efficacy and a strong safety profile when sampling peripheral lung lesions. The synergy of real-time AI aided augmented fluoroscopy and robotic technologies facilitates precise lesion targeting, even within anatomically complex areas. However, follow-up data were not available for non-diagnostic cases; therefore, false-negative rates, sensitivity, and diagnostic accuracy could not be assessed. Given the significant cost associated with this procedure, future prospective studies are warranted to further validate these results and refine patient selection criteria to optimize its clinical application.

IntroductionImmune checkpoint inhibitors (ICIs) have transformed the therapeutic landscape of advanced non-small cell lung cancer (NSCLC), offering durable survival in selected patients. However, ICIs can induce immune-related adverse events (irAEs), among which pneumonitis represents one of the most clinically significant due to its potential severity and diagnostic challenges. Different radiological patterns have been reported, although they may overlap with infectious or radiation-induced lung injury, complicating a timely diagnosis.ResultsWe report two cases of young patients with metastatic NSCLC and high PD-L1 expression who achieved long-term benefit from first-line pembrolizumab but were burdened by severe immune-related pneumonitis. Imaging findings in both patients were suggestive of organizing pneumonia associated with the presence of the so-called “atoll sign” (or reversed halo sign), a radiological feature rarely reported in the context of ICI-related pneumonitis. In both cases, the diagnosis was supported by imaging, microbiological tests, and histology. The subsequent administration of high-dose steroids led to rapid relief of symptoms associated with radiological improvement. Despite pembrolizumab discontinuation, both patients maintained durable systemic disease control.ConclusionThese cases emphasize the importance of recognizing uncommon radiological patterns such as the atoll sign in patients receiving ICIs, highlighting the need for vigilance even after prolonged therapy and underscoring the potential for sustained oncologic benefit despite treatment discontinuation.

BackgroundPrognostic models are crucial for prostate cancer (PCa) treatment decision making at the time of diagnosis, particularly for distinguishing active surveillance (AS) candidates from those requiring curative treatment. While several models exist, their ability to predict metastatic disease—the primary driver of PCa mortality—remains underexplored.MethodsWe analysed the Turin Prostate Cancer Prognostication cohort, which includes 891 unselected PCa patients diagnosed between 2008 and 2013 in Turin, Italy. Three widely used prognostic models—D’Amico, CAPRA, and MSKCC—were updated and compared based on optimism-corrected discrimination and overall prediction error for metastatic PCa (mPCa) within five years of diagnosis, accounting for competing risks. Overall survival was also assessed. Additionally, we investigated whether replacing standard AS eligibility criteria with nomogram-based risk thresholds could better identify patients at low risk of metastasis, maximizing AS uptake while minimising metastatic risk.ResultsThe MSKCC nomogram (optimism-corrected AUCt: 0.81; scaled Brier score: 0.15) outperforming the CAPRA score (AUCt: 0.77; Brier score: 0.11) and the D’Amico classification (AUCt 0.64; Brier score: 0.03) in predicting mPCa. The same ranking was observed for overall mortality prediction. When 95th percentile of MSKCC’s predicted probabilities among patients selected for six different AS protocols was used as a threshold, the proportion of potentially eligible patients increased from 7.8% when UCSF criterion was used to 57.0% without substantially increasing metastatic risk (observed 5-year risk: 1.7%).ConclusionsThe MSKCC nomogram outperformed other models in predicting mPCa and overall mortality. Implementing risk-based AS eligibility thresholds derived from MSKCC could enhance patient selection while facilitating shared decision-making between patients and clinicians.

Background: Approximately 25.0% of metastatic prostate cancer patients harbour DNA damage repair mutations, including BRCA1 and BRCA2, which are actionable targets for poly(ADP-ribose) polymerase (PARP) inhibitors. Accurate detection of BRCA1/2 mutations is critical for guiding targeted therapies, but crucial pre-analytical factors, such as tissue storage duration and DNA fragmentation, drastically affect the reliability of next-generation sequencing (NGS) using real-world diagnostic specimens. Methods: This multicentre study analysed 954 formalin-fixed paraffin-embedded tissue samples from 11 centres, including 559 biopsies and 395 surgical specimens. This study examined the impact of storage duration (<1 year, 1–2 years, and >2 years) and DNA parameters (concentration and fragmentation index) on NGS success rates. Logistic regression and Cox regression analyses were used to assess correlations between these factors and sequencing outcomes. Results: NGS success rates decreased significantly with longer storage, from 87.8% (<1 year) to 69.1% (>2 years). Samples with higher DNA concentrations and fragmentation indexes had higher success rates (p < 0.001). Surgical specimens had superior success rates (83.3%) compared with biopsies (72.8%) due to better DNA quality. The DNA degradation rate was more pronounced in older samples, underscoring the negative impact of extended storage. Conclusions: Timely testing of BRCA1/2 mutations is critical for optimizing the identification of prostate cancer patients eligible for PARP inhibitors. Surgical specimens provide more reliable results than biopsies and minimizing the storage duration significantly enhances testing outcomes. Standardizing pre-analytical and laboratory procedures across centres is essential to ensure personalized treatments and improve patient outcomes.

OBJECTIVES:To evaluate the utility of routine intra-operative frozen section analysis (iFSA) of distal ureteric margins during radical cystectomy (RC) for bladder cancer (BCa), focusing on diagnostic accuracy and oncological outcomes in a high-volume tertiary centre. PATIENTS AND METHODS:We retrospectively identified 1081 patients with BCa treated with RC (2010-2024). Bilateral iFSA of the distal ureters was performed in all cases. Patients were stratified according to final distal ureteric margin status (positive vs negative); if positive, additional resections were attempted intra-operatively. Diagnostic concordance between iFSA and final pathology was calculated. Kaplan-Meier curves were used to assess 5-year upper urinary tract recurrence (UUTR) free-survival, overall survival (OS), and cancer-specific survival (CSS). Univariable logistic regression and multivariable logistic regression (MLR) models identified variables associated with positive margins. RESULTS:Overall, 139 patients (12.9%) had positive distal ureteric margins at iFSA. The sensitivity and specificity of FSA were 98.6% and 99.5%, respectively. In MLR models, hydronephrosis (odds ratio [OR] 1.75, P = 0.014), T3-T4 stage (OR 2.48, P = 0.003), bladder carcinoma in situ (CIS; OR 7.94, P < 0.001) and trigonal tumour location (OR 4.85, P < 0.001) were independently associated with positive distal ureteric margins at iFSA. Positive margins were associated with increased risk of UUTR (5-year UUTR-free survival: 58% vs 78%; P = 0.038), worse OS (5-year OS: 48% vs 67%; P = 0.039), and worse CSS (5-year CSS: 60% vs 75%; P = 0.0018). CONCLUSION:Our study showed that iFSA of distal ureteric margins during RC for BCa provided excellent diagnostic performance and enabled cancer-free anastomosis. Our findings support iFSA, especially in patients with bladder CIS, trigonal tumours, or hydronephrosis, to guide intra-operative decisions and tailor postoperative surveillance.

Non-Small Cell Lung Cancer (NSCLC) is the leading cause of cancer death worldwide. Although immune checkpoint inhibitors (ICIs) have shown remarkable clinical efficacy, they can also induce a paradoxical cancer acceleration, known as hyperprogressive disease (HPD), whose causative mechanisms are still unclear. This study investigated the mechanisms of ICI resistance in an HPD-NSCLC model. Two primary cell cultures were established from samples of a NSCLC patient, before ICI initiation (“baseline”, NSCLC-B) and during HPD (“hyperprogression”, NSCLC-H). The cell lines were phenotypically and molecularly characterized through immunofluorescence, Western Blotting and RNA-Seq analysis. To assess cell plasticity and aggressiveness, cellular growth patterns were evaluated both in vitro and in vivo through 2D and 3D cell growth assays and patient-derived xenografts establishment. In vitro investigations, including the evaluation of cell sensitivity to interferon-gamma (IFN-γ) and cell response to PD-L1 modulation, were conducted to explore the influence of these factors on cell plasticity regulation. NSCLC-H exhibited increased expression of specific CD44 isoforms and a more aggressive phenotype, including organoid formation ability, compared to NSCLC-B. Plastic changes in NSCLC-H were well described by a deep transcriptome shift, that also affected IFN-γ–related genes, including PD-L1. IFN-γ–mediated cell growth inhibition was compromised in both 2D-cultured NSCLC-B and NSCLC-H cells. Further, the cytokine induced a partial activation of both type I and type II IFN-pathway mediators, together with a striking increase in NSCLC-B growth in 3D cell culture systems. Finally, low IFN-γ doses and PD-L1 modulation both promoted plastic changes in NSCLC-B, increasing CD44 expression and its ability to produce spheres. Our findings identified plasticity as a relevant hallmark of ICI-mediated HPD by demonstrating that ICIs can modulate the IFN-γ and PD-L1 pathways, driving tumor cell plasticity and fueling HPD development.

OBJECTIVE:Several studies analyzed the "reprogramming" of germ cell tumors of the testis (GCTT), known to be an epigenetic process that results in the preservation of stem cell features and/or differentiation of GCTT. EZH2 is a methyltransferase involved in the epigenetic regulation of tumors and has become a promising therapeutic target, but few studies have analyzed its expression in GCTT, germ cell neoplasia in situ (GCNIS), and adjacent testis. METHODS:We tested 131, 36, and 29 GCTT components, GCNIS, and adjacent testes, respectively. EZH2 expression was evaluated by H-score and compared between different subgroups by adopting median values and the Fisher exact test. RESULTS:We found that EZH2 was more highly expressed by adjacent testis/GCNIS rather than by GCTT (P < .001), with adjacent testis showing the highest values and being statistically significant compared to GCNIS (P < .001). In adjacent testis, EZH2 expression was mainly detected in spermatocytes (primary and secondary) and spermatids, with scattered positive spermatogonia. Seminoma/embryonal carcinoma showed statistically significantly higher EZH2 expression compared to the other nonseminomatous GCTT (P = .027). CONCLUSIONS:EZH2 is differentially expressed during GCTT reprogramming (adjacent testis [very high levels] → GCNIS [high levels] → seminoma/embryonal carcinoma [moderate levels] → other nonseminomatous GCTT [low/absent levels]), supporting its involvement in the epigenetic regulation for determining the fate of GCTT.

INTRODUCTION:Contemporary prostate biopsy utilizes multiparametric magnetic resonance (MRI) guidance; however, it may fail to identify a non-negligible proportion of men with clinically significant (csPCa). The main objective of this study was to assess the feasibility and diagnostic performance of Prostate Specific Membrane Antigen-Target biopsy (PSMA-TB) to diagnose csPCa in men with negative MRI and high clinical risk of PCa. PATIENTS AND METHODS:Open-label, single-center, nonrandomized, prospective study. INCLUSION CRITERIA:PSA density (PSAd) ≥0.2 ng/ml2 in men with PIRADS 1-2; PSA >10 ng/ml or abnormal digital rectal examination or strong familiar history for PCa or known genetic mutation. Each patients underwent PSMA-PET and transperineal fusion PSMA-TB ± systematic biopsy (SB). RESULTS:Overall, 35 patients were enrolled; 23 (65.7%) men had positive PSMA-PET (PRIMARY score ≥3). Overall, 14 (40%) men had csPCA and 21 (60%) patients had any PCa at PSMA-TB+SB. Only 1 patient (8.3%) with negative PSMA-PET had csPCa (ISUP 3) at SB (92% Negative Predictive Value [NPV]). Fusion PSMA-TB alone detected csPCa in 12 out of 23 (52.2%) patients with positive PSMA-PET; fusion PSMA-TB with concomitant SB increased the detection of csPCa to 56.5% (added value of 4.3%). The sensitivity, specificity, Positive Predictive Value (PPV), NPV and AUC of PSMA-TB+SB were 93%, 57%, 59%, 92% and 0.75 for detection of csPCa and 91%, 79%, 86%, 95% and 0.84 for detection of any PCa, respectively. The main limitation of this study is its small sample size. CONCLUSIONS:Fusion PSMA-TB is technically feasible and may improve the detection of csPCa in patients with negative MRI.

Introduction Prostate cancer diagnosis and treatment planning depend on accurate histopathological assessment of needle biopsies, particularly through the Gleason scoring system. The inherently subjective nature of the grading creates variability between pathologists, potentially resulting in suboptimal patient management decisions. These reproducibility challenges extend beyond Gleason scoring to encompass other critical diagnostic and prognostic markers, including cancer volume quantification and detection of cribriform morphology patterns and perineural invasion. Artificial intelligence (AI) applications in digital pathology have emerged as promising solutions for enhancing diagnostic consistency and accuracy, with recent research demonstrating that automated systems can match expert-level performance in prostate biopsy evaluation. Nevertheless, comprehensive validation studies have revealed concerning limitations in model generalisability when deployed across different clinical environments and patient populations. Recent systematic reviews revealed widespread risk-of-bias limitations and insufficient external validation in AI diagnostic studies, highlighting critical needs for accumulated evidence supporting generalisability before clinical implementation. Rigorous external validation with preregistered protocols using independent datasets from diverse clinical settings remains essential to establish the reliability and safety of AI-assisted prostate pathology systems.Methods and analysis This study protocol establishes a framework for the retrospective external validation of an AI system developed for prostate biopsy assessment, to be conducted on the case-control samples of the National Prostate Cancer Register of Sweden, ProMort study (1998-2015). The primary aim is to evaluate the AI model’s diagnostic accuracy and Gleason grading performance using completely independent datasets separate from any model development or previously used validation cohorts. The diversity of the validation samples, spanning multiple geographic regions, temporal collection periods and reference standards, allows evaluation of model robustness across varied clinical contexts. Secondary aims encompass evaluating AI performance in cancer length estimation and detection of cribriform patterns and perineural invasion. This protocol delineates procedures for data collection, reference standard clarification and prespecified statistical analyses, ensuring comprehensive validation and reliable performance assessment. The study design conforms to established reporting guidelines Checklist for Artificial Intelligence in Medical Imaging (CLAIM) and Standards for Reporting Diagnostic Accuracy Studies using Artificial Intelligence (STARD-AI), and recognised best practices for AI validation in medical imaging.Ethics and dissemination Data collection and usage were approved by the Swedish Regional Ethics Review Board and the Swedish Ethical Review Authority (permits 2012/1586-31/1, 2016/613-31/2, 2019-01395, 2019-05220). The study adheres to the Declaration of Helsinki principles, and findings will be made available in open access peer-reviewed publications.

OBJECTIVE To evaluate the occurrence and the oncological predictive value of cribriform growth and/or intraductal carcinoma (CR/IDC) in patients with ISUP grade group (GG) 2 prostate cancer (PCa) at radical prostatectomy (RP) with and without synchronous nodal metastases in a multicenter, international cohort. METHODS We identified 1060 patients who underwent RP with ISUP GG2 PCa at histopathology and a pelvic lymph node dissection from 3 tertiary referral centers. Of these, 79 (7.4%) had pN1 disease. Case-control matching was performed using the initial prostate-specific antigen (iPSA) value, pTstage, age, surgical margin status, and referral center as matching variables to compare histopathological characteristics and oncological outcomes between pN1 and pN0 patients. The predictive value of CR/IDC for biochemical recurrence-free survival (BCRFS), defined as the interval between RP and a PSA of >= 0.2 ng/mL, and radiological recurrence-free survival (RRFS), defined as the interval between RP and an RR, was evaluated using Cox regression analysis. RESULTS After case-control matching, 106 patients were included (ie, 53 cases and 53 controls). CR/IDC was significantly more common in pN1 than pN0 RP specimens (100% vs 51%, P < .001). In pN0 patients, CR/IDC positivity was not associated with BCRFS (hazard ratio [HR] = 0.90, 95% CI 0.32-2.55, P = .842) or RRFS (HR 2.45, 95% CI 0.45-13.34, P = .299). pN1 CR/IDC-positive PCa was associated with adverse BCRFS (HR = 2.93, 95% CI 1.26-6.83, P = .013) and RRFS (HR = 9.19, 95% CI 2.11-40.04, P = .003) in multivariable Cox regression analysis. CONCLUSION In ISUP GG2 PCa, CR/IDC strongly correlates with synchronous nodal metastases, the latter being associated with adverse outcomes. UROLOGY 199: 104-111, 2025. (c) 2025 Published by Elsevier Inc.

BACKGROUND:Immune-based combinations have transformed first-line treatment for metastatic renal cell carcinoma (mRCC), but reliable biomarkers for patient selection remain elusive. Chromosome 3p mutations (e.g., VHL, PBRM1, SETD2, BAP1) have shown inconsistent prognostic and predictive value. This retrospective study assessed the prognostic impact of tissue-based biomarkers, focusing on 3p mutations in mRCC. PATIENTS AND METHODS:A single-center retrospective analysis included mRCC patients treated with immunocombinations or tyrosine kinase inhibitors (TKIs). We evaluated mutations in 14 genes, including VHL, PBRM1, SETD2, and BAP1. Primary endpoints were overall survival (OS) and progression-free survival (PFS). Prognostic factors were analyzed using univariate and multivariate models. RESULTS:Among the included 38 patients, the most common mutations were in VHL (45 %), PBRM1 (42 %), and SETD2 (26 %), with these latter more frequent in males (p = 0.012). Most patients (74 %) received immune-based combinations; 26 % received TKIs. SETD2 mutations were associated with primary refractoriness (30 %). Median OS was not reached; brain metastases (p = 0.001) and BAP1 mutations (p = 0.025) predicted worse OS, while PBRM1 mutations trended toward improved OS (p = 0.845). Median PFS was 14.1 months. Higher tumor grade (p = 0.038) and worse ECOG PS (p = 0.008) negatively impacted PFS, while 3p mutations showed no significant effect on PFS. CONCLUSIONS:ECOG PS and brain metastases were confirmed as poor prognostic factors. VHL and PBRM1 mutations may suggest a better prognosis, while SETD2 and BAP1 mutations portend worse outcomes. Larger studies are needed to confirm these findings.

The rapid evolution of digital pathology has enabled large-scale data acquisition, driving sophisticated clinical research and advancing the development of AI-driven tools. These innovations have also revolutionised histopathological slide review, especially the annotation step (i.e. the process of marking specific areas of interest on glass-mounted tissue samples to add relevant clinical information) by digitising the process, enhancing precision and efficiency, and facilitating collaboration. However, currently available open-source annotation tools typically employ single-label approaches that provide a flat representation of whole-slide images (WSI), limiting their ability to capture the complexity of the diagnosis-significant elements in a detailed and structured way. Furthermore, the difficulty of strictly following precise review protocols and lack of provenance tracking during annotation processes can result in high variability and limit reproducibility and reusability of the collected data. In this work we present the CRS4 Digital Pathology Platform (CDPP), an open-source system for research studies that manages WSI collections and focuses on high-quality, structured annotations, gathered according to well-defined protocols. Its main features include: (1) structured, multi-label morphological and clinical image annotation; (2) support for controlled but customisable annotation protocols; (3) dedicated annotation tools to facilitate enhanced accuracy, efficiency and consistency in the annotation process; and (4) workflow-based computational analysis with integrated provenance tracking. We show how the platform has successfully supported three different studies, demonstrating the CDPP’s ability to assist pathologists in the generation of high-quality annotated datasets, also suitable for reuse, in the context of digital pathology research.

Supplementary Figure S8 shows migration assays of MYC-CaP cells treated with lactate or LDHA inhibitor FX11.

Excel file containing 19 supplementary tables

Supplementary Figure S1 shows the systemic effects of HFD.