Artificial intelligence (AI) for pathology is accelerating cancer diagnostics. For global adoption, studies in currently under-represented populations - where the potential gains from AI may also be greatest - are needed. We present the first external validation study in a Middle Eastern setting, focusing on diagnosis and Gleason grading of prostate cancer. We digitised 339 prostate glass slides from Kurdistan, Iraq, representing a consecutive series of 185 patients from 2013 to 2024. Each slide was scanned on three digital pathology scanners, generating 1,017 whole-slide images. Task-specific and foundation AI models demonstrated AI-pathologist grading concordance similar to pathologist-pathologist concordance (Cohen's quadratically weighted kappa 0.801 vs. 0.799). Cross-scanner concordance was high (kappa > 0.90) for all AI models across three scanners. The demonstrated robust performance on a low-cost compact scanner could enable validation studies in non-digitalised settings and cost-effective AI adoption in laboratories with limited sample volumes. The publicly shared dataset supports further research to facilitate broader adoption of AI pathology.

Gastric dysplasia comprises a heterogeneous group of non-invasive epithelial neoplasms with distinct clinicopathological and biological characteristics. This review focuses on surface epithelium-derived lesions, including intestinal-type, foveolar-type, and serrated dysplasia/adenoma. Conventional low-grade intestinal-type dysplasias/adenomas arise mainly in atrophic and metaplastic mucosa associated with Helicobacter pylori infection or, much less frequently, autoimmune gastritis. Small, pale, flat-elevated lesions with orderly small-intestinal differentiation generally behave indolently, whereas large size, redness, depression, nodularity, high-grade histology, and abnormal p53 expression are associated with an increased risk of progression. Foveolar-type dysplasia/adenoma is particularly heterogeneous. Flat-elevated foveolar-type adenoma, raspberry-type adenoma, and dysplasia arising in fundic gland polyps represent distinct dysplasia subtypes with indolent behavior. In contrast, other foveolar-type lesions are biologically heterogeneous; some are associated with microsatellite instability-high or Epstein-Barr virus-associated gastric carcinoma, whereas others may represent the superficial component of poorly differentiated or deceptively well-differentiated adenocarcinoma. Gastric serrated dysplasia is rare but is frequently associated with invasive carcinoma and should be regarded as a high-risk lesion requiring complete resection. Accurate diagnosis therefore requires integration of histological findings with endoscopic appearance and the background gastric mucosa. The principal role of pathological diagnosis is to distinguish dysplastic lesions with indolent behavior from those carrying a non-negligible risk of progression to invasive carcinoma, thereby guiding optimal clinical management.

Well-differentiated pancreatic neuroendocrine tumors measuring ≤ 2 cm (small PanNETs) are usually non-functional and most are discovered incidentally. While most are indolent, some behave aggressively and metastasize. Alternative lengthening of telomeres (ALT) has shown promise as a predictive biomarker of recurrence in recent studies. We designed a multi-institutional study to validate the predictive ability of ALT in small PanNET to inform patient selection for operative versus non-operative management. We identified 172 cases of small PanNET diagnosed from 2001 to 2021 in four different institutions. ALT fluorescence in situ hybridization (FISH) could be performed in 133 cases and was compared with clinical outcomes. Among those with FISH, 44.4% were females with a mean age of 65.4 years (range 31-88). Twenty-one cases (15.8%) were ALT-positive. Over a median follow-up of 48 months, a total of 19 progressions (metastasis, recurrence, increase in size, or death) were recorded. ALT positivity was significantly associated with worse progression-free survival [PFS]. In the ALT-positive group, a total of 9 cases showed progression compared to 10 in the ALT-negative group (5-year PFS: 70.2% vs. 88.8%, hazard ratio: 4.69, p = 0.002). In a separate analysis using matched cases, ALT chromogenic in situ hybridization (CISH), demonstrated high concordance with ALT FISH (95.5% for positive and 91.7% for negative cases) enabling cross-platform validation. Unlike FISH, CISH can be interpreted using standard bright field microscopy, facilitating broader implementation. These findings support further evaluation of ALT testing, particularly via CISH, as a potentially scalable and accessible biomarker, while recognizing that formal analytic validation and confirmation in unselected cohorts are required before routine clinical implementation.

The current classification of pulmonary neuroendocrine neoplasms (NENs) is primarily based on histomorphological criteria, which do not fully capture their underlying biological heterogeneity. This limitation is particularly evident among pulmonary carcinoids, a subgroup characterized by highly variable clinical behaviour and prognostic outcomes despite similar morphological features.Angiogenesis is a hallmark of NENs, yet the role of the vascular endothelial growth factor (VEGF) pathway in pulmonary NENs remains incompletely defined.In this multicentre retrospective study, 109 pulmonary NENs (67 typical carcinoids, 22 atypical carcinoids, and 20 high-grade NENs) were analysed. VEGF, VEGFR1, VEGFR2, and VEGFR3 expression was assessed by immunohistochemistry and correlated with clinicopathological features and outcome.VEGF pathway components showed subtype-specific expression patterns. VEGFR1 was reduced in high-grade tumours and displayed a bimodal distribution, while VEGFR2 and VEGFR3 were higher in atypical carcinoids and high-grade tumours. Survival and recurrence were driven by histological subtype, whereas VEGF pathway expression lacked independent prognostic significance.The higher expression of VEGFR2 and VEGFR3 observed in atypical carcinoids and high-grade neuroendocrine carcinomas suggests a possible involvement of the VEGF signalling pathway in tumour biology. These findings contribute to the biological characterization of pulmonary NENs and provide a rationale for future studies investigating whether VEGF pathway markers may have predictive relevance for anti-angiogenic therapies.

The full morphologic spectrum of divergent differentiations in urothelial carcinoma is still unknown. Herein, we describe two cases of pilomatrix-like high grade carcinomas (PiMLC) involving bladder. Case #1 was a 55-year-old male with a bladder tumor with basaloid nests, geographic necrosis, and shadow cells. The tumor cells were positive for LEF1, β-catenin (nuclear/cytoplasmic staining), CK5/6, CD10, BerEP4, p63 (focal-to-patchy) and AMACR (patchy), and negative for CK7, GATA3, CDX2, INSM1, chromogranin, synaptophysin, p40, CK20, calponin, CD117, S100, smooth muscle actin, p16 and high-risk HPV in-situ hybridization. PD-L1 and Her2 were negative. Trop-2 immunostain showed modest expression (H-score: 75). There were rare microscopic foci of surface disease with a similar basaloid appearance that were also positive for LEF1 and negative for GATA3. Genomic profiling of the tumor revealed an APC mutation, suggesting β-catenin/Wnt pathway dysregulation. Additionally, mutations in TERT, CDKN2A, and TP53 were present. The patient was transitioned to hospice shortly after initiating chemotherapy. Case #2 was a 78-year-old male with a large bladder mass with basaloid nests and abundant shadow cells. Conventional UC was focally present. The tumor cells were positive for p63, β-catenin (nuclear/cytoplasmic staining), CK14 (focal), and CK20 (focal), and negative for GATA3, Uroplakin, NUT, and Her2. Genomic profiling of the tumor revealed a CTNNB1 mutation, suggesting β-catenin/Wnt pathway dysregulation. Additionally, mutations in RB1, KMT2A, and TP53 were present. Overall, these two cases describe PiMLC of urinary bladder with evidence supporting novel pilomatrix-like divergent differentiation from urothelial carcinoma.

EIF1AX alterations occur across benign, borderline, differentiated, and high-grade thyroid neoplasia, requiring contextual interpretation. PubMed/MEDLINE, Scopus, and Embase were searched through 6 May 2026; complementary genomic, citation, and supplementary-data searches continued through 19 July 2026. Analyses addressed prevalence in strict Bethesda III/IV nodules, tissue-verified outcomes in EIF1AX-positive Bethesda III/IV nodules, and prevalence in WHO- or Turin-defined poorly differentiated thyroid carcinoma (PDTC) and histology-defined anaplastic thyroid carcinoma (ATC) cohorts not molecularly preselected. Proportions were pooled with binomial-normal generalized linear mixed models; study intervals were exact Clopper-Pearson intervals and heterogeneity was τ[2] on the logit scale. Among 43 reports, 13 contributed to a primary estimate and 17 to a primary or sensitivity analysis. EIF1AX prevalence in three strict Bethesda III/IV cohorts was 4.75% (92/1935; 95% CI, 3.89%-5.80%; τ[2]=0.000). Malignant histology occurred in 50/117 tissue-verified nodules (40.77%; 95% CI, 24.59%-59.24%; τ[2]=0.413). Including noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP)/borderline outcomes and compatible Desai and Connelly data yielded 64/124 (48.16%; 95% CI, 30.69%-66.10%; τ[2]=0.497). Prevalence was 6.15% in PDTC (11/136; 95% CI, 2.04%-17.12%; τ[2]=0.455) and 15.06% in ATC (25/162; 95% CI, 9.19%-23.70%; τ[2]=0.154). EIF1AX is uncommon in indeterminate cytology and is not a stand-alone rule-in marker. Interpretation requires variant class, co-alterations, assay, verification, and histologic context.

Sequential lymphoid neoplasms are rare and diagnostically challenging, with a variable clonal relationship. Here we present the case of a 59-year-old woman with mediastinal grey zone lymphoma (MGZL), who, shortly after achieving complete remission, developed a subsequent cutaneous neoplasm. Comparative molecular studies were performed using high-throughput sequencing (HTS), which demonstrated that both neoplasms shared high-allele-frequency EP300:p.P2333L and PTPRK:p.R532K variants. In addition, each neoplasm harboured private, mutually exclusive variants-NFKBIA:p.R245Sfs*39 and BTG2:c.142+5G>C in the MGZL and JAK1:p.Q562* and JAK1:p.G1097D in the skin. Integrating the clinical course with HTS, we diagnosed lymphomatoid papulosis and excluded relapse of the MGZL. Interestingly, the shared variants were likely germline rather than clonal, suggesting they created a permissive background predisposing to two independent transforming events along the B- and T-cell lineages. Molecular studies may help to reveal the genetic basis of composite lymphomas and to resolve diagnostically discordant, complex presentations.

Deep learning has shown remarkable success in cytopathology, particularly in cervical cancer screening and thyroid nodule classification. Yet, most models remain "black boxes," offering deterministic predictions without indicating uncertainty or explaining their reasoning, limiting clinical adoption. This review systematically examines uncertainty quantification (UQ) and explainable AI (XAI) techniques for trustworthy deep learning in cytology-the first to do so in this domain. We review state-of-the-art methods such as Bayesian neural networks, Monte Carlo dropout, and deep ensembles for uncertainty estimation, alongside Grad-CAM, attention mechanisms, and surrogate models for interpretability. Applications span cervical, thyroid, urinary, and hematological cytopathology. UQ methods help identify model limitations, flagging cases that need human review, while studies show improved accuracy for high-confidence predictions and better detection of misclassified cases. Visual explanation tools, especially Grad-CAM, highlight diagnostically relevant cellular features that align with pathologists' reasoning. Integrating UQ and XAI fosters effective human-AI collaboration, enabling systems to communicate both confidence levels and diagnostic rationale. This addresses the clinical trust gap, positioning AI as an assistive colleague rather than a black box, ultimately enhancing diagnostic accuracy and efficiency in cytopathology workflows.

Adrenocortical carcinoma (ACC) is a rare endocrine malignancy with heterogeneous outcomes after resection. We conducted a retrospective single-institution cohort study of 65 resected primary ACCs from 2010 to 2025. Clinicopathologic features, tumor-type-specific histologic scoring, immunohistochemistry for p53, β-catenin, and Ki-67, clinical molecular testing results, and outcomes were analyzed. Survival was evaluated using Kaplan-Meier methods and Cox models adjusted for pT category. Increased mitotic activity (> 5 mitoses per 50 high-power fields) was present in 41/55 tumors (74.5%). Clinical molecular testing was performed in 32 tumors; 12 (37.5%) had at least one pathogenic or likely pathogenic (P/LP) alteration, most commonly involving TP53 (6/12) and CTNNB1 (3/12). P/LP alteration positivity was associated with worse overall survival (OS) after pT adjustment (HR, 4.67; 95% CI, 1.26-17.26; p = 0.021). Nuclear β-catenin expression was also associated with worse OS after pT adjustment (HR, 3.23; 95% CI, 1.19-8.79; p = 0.022). Increased mitotic activity remained associated with worse OS after pT adjustment (HR, 4.62; 95% CI, 1.07-19.91; p = 0.040), whereas its association with disease-specific survival did not reach statistical significance after pT adjustment (HR, 7.44; 95% CI, 0.98-56.21; p = 0.052). Ki-67 ≥ 10%, aberrant p53 staining, tumor necrosis, and angioinvasion were not significantly associated with survival after pT adjustment. Oncocytic ACCs had less frequent increased mitotic activity than conventional ACCs (44.4% vs. 79.1%; p = 0.048). Mitotic activity, nuclear β-catenin expression, and clinically reported P/LP alterations may provide complementary prognostic information in resected primary ACC. Larger cohorts with standardized pathology review and uniform molecular profiling are needed.

Pancreatic ductal adenocarcinoma (PDAC) remains highly lethal; however, early-stage cases with improved survival are increasingly detected, underscoring the need for more standardized and clinically applicable histological grading in routine diagnostic practice. Based on 139 resected PDACs without neoadjuvant therapy, including 41 pT1 tumors, we propose a modified CAP-based grading system (mCAP). This system assesses the overall proportion of non-tubular components (NTC) and dense non-tubular clusters (DNTC), the latter defined as ≥ 30 non-tubular clusters within a 0.785 mm[2] area. DNTC was incorporated to capture focal dense non-tubular morphology not adequately reflected by overall NTC proportion alone. Tumors were classified as G1 (< 10% NTC and DNTC-negative), G2 (10-<50% NTC or < 10% NTC with DNTC positivity), or G3 (≥ 50% NTC). In Kaplan-Meier and log-rank analyses, mCAP showed prognostic stratification for both disease-free survival (DFS) and overall survival (OS). In multivariable Cox analysis for pT1c-pT3 tumors, mCAP G2/G3 remained independently associated with worse prognosis: the adjusted hazard ratios for mCAP G2/G3 versus G1 were 5.23 for DFS (95% confidence interval [CI], 2.58-10.61; P < 0.001) and 3.06 for OS (95% CI, 1.66-5.65; P < 0.001). In grade-only Cox models, mCAP showed the lowest Akaike information criterion (AIC) values for both DFS and OS among the three grading systems, suggesting more favorable model fit for mCAP. In the interobserver reproducibility study, unweighted Fleiss' kappa coefficients were 0.287 for WHO, 0.500 for CAP, and 0.531 for mCAP. Overall, mCAP may refine morphology-based risk stratification in resected PDAC.

Hybrid/collision odontogenic tumors are a rare and controversial phenomenon not formally recognized as a distinct category in the WHO Classification of Head and Neck Tumors and their diagnosis relies mainly on histomorphology. This report presents a unique mandibular odontogenic neoplasm with molecular and morphologic features in support of a true hybrid/collision process demonstrating distinct ameloblastic fibroma and ameloblastoma components.

Lung neuroendocrine neoplasms (NENs) range from low-grade carcinoids to high-grade aggressive neuroendocrine carcinomas (NECs). While studies have identified molecular subtypes of NEC driven by lineage-defining transcription factors, the molecular classification of lung carcinoids (LCs) remains poorly understood. Here, we investigated the molecular and clinicopathological characteristics of LC, focusing on hepatocyte nuclear factor 4α (HNF4α), a transcription factor expressed in certain NECs. Using a publicly available transcriptomic dataset comprising 30 carcinoid tumors, we identified a distinct HNF4A-high LC subtype, in which HNF4A expression was almost mutually exclusive with NK2 Homeobox 1 (NKX2-1), which encodes thyroid transcription factor-1 (TTF-1), and Achaete-Scute family BHLH transcription factor 1 (ASCL1), showing a trend toward central/endobronchial locations. Immunohistochemical analysis of 15 surgically resected LCs revealed that HNF4α-positive LCs accounted for 46.7% (7/15) and HNF4α-negative LCs and tumorlets were mostly TTF-1-positive. RNA sequencing of representative HNF4α-positive and -negative cases demonstrated that the HNF4α-positive subtype possesses a distinctive molecular signature. Differentially expressed genes identified across two independent transcriptome datasets revealed that HNF4α-positive/HNF4A-high and HNF4α-negative/HNF4A-low LCs were enriched for genes related to the digestive and pulmonary neuroendocrine systems, respectively, suggesting distinct cells of origin. Furthermore, somatostatin receptor 2 (SSTR2) was significantly upregulated in the HNF4α-positive/HNF4A-high subtype, as validated by immunohistochemistry. Taken together, HNF4α-positive LC represents a unique subtype characterized by a specific neuroendocrine phenotype, mutually exclusive with TTF-1 and ASCL1 expression, and SSTR2 may serve as a potential therapeutic target for this subtype.

We report the clinical, morphological, ultrastructural, and molecular features of a gastroblastoma diagnosed in a 27-year-old male describing some unreported findings that can help to better understand this very rare tumor. The clinical course lasting 28 years confirms the indolent nature of gastroblastoma, although the disease was metastatic to local lymph nodes and peritoneum at the time of diagnosis. The tumor displayed a predominant monophasic appearance, being composed of epithelial cells with only very rare spindle mesenchymal cells, confirmed at ultrastructural examination. Thorough immunohistochemical investigation demonstrated coexpression of cytokeratins (AE1/AE3, CAM5.2, CK8/18), podoplanin (D2-40), CD56, and NSE in absence of synaptophysin, chromogranin A, DOG1, CD117, CD34, SMA, desmin, vimentin, and S100. In addition, the expression of somatostatin receptor 2 A associated with positive octreoscan and [68]Ga-DOTATOC PET imaging may suggest the use of this nuclear medicine approach for the staging of the tumor and to select patients for possible therapy with somatostatin analogues. The tumor was mismatch repair proteins proficient and did not show any mutation in all 60 genes investigated. Interestingly, targeted RNA sequencing identified a previously unreported TAC3::GLI1 fusion that expands the spectrum of genetic aberrations of gastroblastoma.