Nassar, Aziza

Abstract Large language models (LLMs) and vision‐language models represent a fundamentally different category of artificial intelligence (AI) compared to prior image analysis approaches in digital pathology, which have largely been based on convolutional neural network architectures. This review from the American Society of Cytopathology Clinical Practice Committee examines the current evidence for LLM and vision‐language model applications in cytopathology, including structured reporting, diagnostic assistance, quality control, education, and workflow integration. The distinction between applications with preliminary evidence and those that remain hypothetical is described. A detailed assessment of the challenges that must be addressed before clinical deployment, including hallucination risk, limited explainability, bias, data privacy, validation gaps, and infrastructure barriers is discussed. A review of the regulatory landscape in the United States and European Union as it applies to AI‐enabled software as a medical device is provided. Recommendations addressing cytopathology‐specific benchmarks, multi‐institutional validation, transparent governance, and incremental deployment beginning with low‐risk applications are suggested. In the current environment, LLMs have the potential to augment cytopathology practice, but responsible adoption requires rigorous validation and sustained collaboration among cytopathologists, AI researchers, and regulatory bodies.

ABSTRACT Background Transvenous lead extraction (TLE) is a commonly performed and generally safe procedure, though procedural complexity and risk can vary based on clinical and procedural factors. Objective This study aims to identify clinical and procedural factors associated with histologic evidence of subclinical vascular injury, defined by elastin positivity on extracted leads and its clinical significance. Methods Patients undergoing TLE with a lead dwell time of ≥ 1 year were included. Extracted leads with tissue were analyzed histologically using Verhoeff–Van Gieson staining to assess the presence of elastin. Demographic data, procedural and lead characteristics, and indication for extraction were obtained. Established extraction risk scores were calculated to determine efficacy in predicting subclinical vascular injury. Results The cohort included 49 patients with 92 leads extracted. A total of 25 patients (51%) had elastin‐positive leads. Complete procedural success occurred in 95.9% of cases. The multivariate regression model did not identify any independent predictors of elastin presence, including age (OR = 1.00; 95% CI 0.95–1.05; p = 0.89), female gender (OR = 3.04; 95% CI 0.67–13.65; p = 0.14), lead dwell time (OR = 1.13; 95% CI 0.98–1.29; p = 0.07). Additionally, no scoring system, such as the CCF, EROS, MB, SAFeTY, and LECOM, predicted the presence of elastin. Conclusion Subclinical venous injury was common following TLE, yet did not predict adverse clinical outcomes, reinforcing that TLE performed in experienced centers carries a favorable safety profile. Future studies in larger cohorts are needed to further characterize the clinical utility of subclinical vascular injury.

564 Background: HER2+ ILC is rare (~5% of ILC), with limited data on long-term adjuvant trastuzumab outcomes. We investigated late recurrence and immune landscapes of HER2+ ILC vs invasive ductal carcinoma (IDC) in the phase III NCCTG N9831 trial (NCT00005970) and a real-world cohort. Methods: We evaluated 3304 early-stage HER2+ breast cancer patients (pts) enrolled in N9831 (122 ILC, 3182 IDC). Recurrence-free survival (RFS) and overall survival (OS) were estimated (Kaplan-Meier) and compared using multivariable Cox regression adjusting for age, tumor size, grade, hormone receptor (HR), and nodal status. Findings were validated in a real-world cohort of HER2+ pts (Caris Life Sciences CODEai platform). Immune microenvironments were profiled by WTS deconvolution (Quantiseq). Real-world overall survival was calculated from trastuzumab start to last contact. Results: In N9831, ILC pts were older (median 54 vs 49 y), more ER+ (74.6% vs 51.4%), and had more N3 disease (25.4% vs 12.7%) compared to IDC. ILC pts derived a comparable magnitude of benefit from adjuvant trastuzumab compared to the overall population (HR 0.58 vs 0.67). However, long-term outcomes diverged: while the 5-year RFS was similar (77.1% vs 82.7%), ILC pts experienced a steep decline between years 5-15 (77.1% to 57.7%; ~20% decline) vs IDC (82.7% to 72.7%; ~10% decline) after multivariable Cox analysis (Table). In the real-world early/advanced cohort (n=1328; 26 ILC vs 1302 non-ILC [subset IDC n=595]) treated with trastuzumab-based therapy, ILC pts had worse mOS (33.7 vs 68.6 m, HR 2.01, 95%CI 1.00-4.07, p=0.04) compared to non-ILC. However, there was no statistically significant difference when compared specifically to IDC (HR 1.77, 95% CI 0.87-3.61, p=0.11). ILC had: i) numerically lower tumor mutation burden (TMB)-high (4.2% vs 9.5%, p=0.38); ii) higher infiltration of B cells, Tregs, and M2 macrophages (p<0.05); iii) upregulated MHC class II expression (HLA-DOA, p<0.05); and iv) downregulated proliferative and DNA repair pathways (E2F, G2M, MYC; FDR <0.25) compared to non-ILC. Conclusions: HER2+ ILC carries late recurrence risk despite initial trastuzumab benefit. The unique immune landscape (Treg/M2 macrophage enrichment and lower proliferative signatures) distinguishes HER2+ ILC from IDC, highlighting the need for tailored strategies. Long-term RFS and OS in HER2-positive ILC vs IDC (N9831 trial). Population N 5-year% (95% CI) 10-year% (95% CI) 15-year% (95% CI) ILC 122 RFS 77.13% (69.9, 85.1) OS 84.04% (77.7, 90.9) RFS 70.57% (62.7, 79.4) OS 76.0% (68.7, 84.2) RFS 57.67% (44.2, 75.1) OS 65.22% (56.6, 75.2) IDC 3182 RFS 82.71% (81.4, 84.1) OS 90.18% (89.1, 91.2) RFS 77.45% (75.9, 78.9) OS 81.81% (80.4, 83.2) RFS 72.68% (70.77, 74.6) OS 76.68% (75, 78.4)

Foundation models are reshaping computational pathology by enabling scalable task-agnostic representations of histopathological whole-slide images (WSIs). Unlike earlier task-specific deep learning systems, pathology foundation models (PFMs) leverage massive whole-slide image repositories and self-supervised Vision Transformer architectures to achieve broad generalization and few-shot adaptability. Their evolution reflects a shift from weakly supervised approaches such as Clustering-Constrained Attention Multiple Instance Learning (CLAM) and hierarchical architectures such as Hierarchical Image Pyramid Transformer (HIPT) to large-scale efforts including foundation models, UNI, Virchow, Phikon, CONtrastive learning from Captions for Histopathology (CONCH), GigaPath, H-Optimus, Transformer-Based Pathology Image and Text Alignment Network (TITAN), and the Mayo Clinic Atlas. These models demonstrate impressive performance across diagnostic and prognostic benchmarks while also opening pathways for multimodal integration with genomics and clinical data. Yet significant barriers remain including inconsistent generalization across institutions, interpretability lagging behind clinical needs, and slow integration into routine laboratory workflows. Certain domains of anatomic pathology such as cytopathology, transplant pathology, frozen sections, and rare tumor subtypes remain particularly resistant to current models. Here, we review the development of PFMs, critically evaluate their strengths and limitations, and outline priorities for their safe and effective clinical translation. We argue that the next phase of PFM development will depend on rigorous benchmarking, pathologist-in-the-loop deployment, and multimodal fusion ensuring these models evolve from research tools into clinically robust systems.

Anal squamous cell carcinoma (ASCC) is a rare malignancy of the lower gastrointestinal tract, with distant metastases typically involving the liver and lungs. Metastasis to the kidneys is uncommon, and only one prior case has been reported in the literature. Herein, we report a rare presentation of a patient with biopsy-confirmed metastatic ASCC presenting as bilateral renal lesions. We then provide a review of the literature for rare metastatic presentations of ASCC, highlighting all the cases described in the literature. Clinicians should maintain a high index of suspicion for unusual metastatic sites, perform targeted imaging and biopsy when indicated, and consider systemic therapies to optimize outcomes in rare metastatic presentations.

Artificial Intelligence (AI) has the potential to revolutionize cytopathology by enhancing diagnostic accuracy, efficiency, and accessibility. However, the implementation of AI in this field presents significant challenges and opportunities. This review paper explores the current landscape of AI applications in cytopathology, highlighting the critical challenges, including data quality and availability, algorithm development, integration and standardization, and clinical validation. We discuss challenges such as the limitation of only one optical section and z-stack scanning, the complexities associated with acquiring high-quality labeled data, the intricacies of developing robust and generalizable AI models, and the difficulties in integrating AI tools into existing laboratory workflows. The review also identifies substantial opportunities that AI brings to cytopathology. These include the potential for improved diagnostic accuracy through enhanced detection capabilities and consistent, reproducible results, which can reduce observer variability. AI-driven automation of routine tasks can significantly increase efficiency, allowing cytopathologists to focus on more complex analyses. Furthermore, AI can serve as a valuable educational tool, augmenting the training of cytopathologists and facilitating global health initiatives by making high-quality diagnostics accessible in resource-limited settings. The review underscores the importance of addressing these challenges to harness the full potential of AI in cytopathology, ultimately improving patient care and outcomes.

Abstract Lung transplant recipients are at higher risk of developing COVID-19 infection compared to other solid organ transplants. The risk further increases in the unvaccinated patients. We present a case of a 43-year-old male who underwent bilateral sequential lung transplantation for pulmonary alveolar microlithiasis (PAM) and had an uneventful recovery. However, two years post-transplantation, the patient developed chronic lung allograft dysfunction (CLAD) with bronchiolitis obliterans syndrome and two episodes of COVID-19 infection. During the second episode of COVID-19 infection, the patient developed sepsis and multi-organ dysfunction ultimately resulting in death. Our case report highlights the increased susceptibility of PAM patients’ post-lung transplant to COVID-19 infection. Continuous follow-up of PAM patients’ post-lung transplantation is necessary to prevent unfavorable outcomes.

AbstractBackgroundCytology is extremely important for diagnosis of lung carcinoma and the Papanicolaou Society of Cytopathology (PSC) had proposed a new classification system for respiratory cytology for better communication between physicians and better patient management. The objective of this study is to analyze our samples in accordance with this classification and to evaluate the diagnostic accuracy of various cytologic techniques and to assess the risk of malignancy.MethodsEight hundred and twenty respiratory cytology specimens (FNA, BAL, washing, brushing, sputum) collected between 2019 and 2022 were classified according to the PSC system and the risk of malignancy was assessed for each category using follow‐up surgical samples. Sensitivity, specificity, and accuracy rates were determined based on a categorial approach, according to a similar study.ResultsThe data of 820 respiratory cytology specimens from 576 patients were analyzed. 2.6% of these were non‐diagnostic, 64.1% were NM, 5% were AC, 0.4% were N‐B‐LG, 4% were SM and 23.9% were ML. The risk of malignancy for each diagnostic category were: 42.8% for non‐diagnostic, 31.2% for NM, 43.9% for AC, 87.9% for SM, 94.3% for ML. Sensitivity and specificity was calculated using only the malignant cases considered as positive tests and was 45.57% and 97.34% respectively.ConclusionOur results correlated with the PSC system, and it was considered useful in clinical practice. However, more studies should be performed to evaluate the usefulness of this system. The ROMs of each category and the impact of different techniques should be further studied.

Transplant pathology plays a critical role in ensuring that transplanted organs function properly and the immune systems of the recipients do not reject them. To improve outcomes for transplant recipients, accurate diagnosis and timely treatment are essential. Recent advances in artificial intelligence (AI)-empowered digital pathology could help monitor allograft rejection and weaning of immunosuppressive drugs. To explore the role of AI in transplant pathology, we conducted a systematic search of electronic databases from January 2010 to April 2023. The PRISMA checklist was used as a guide for screening article titles, abstracts, and full texts, and we selected articles that met our inclusion criteria. Through this search, we identified 68 articles from multiple databases. After careful screening, only 14 articles were included based on title and abstract. Our review focuses on the AI approaches applied to four transplant organs: heart, lungs, liver, and kidneys. Specifically, we found that several deep learning-based AI models have been developed to analyze digital pathology slides of biopsy specimens from transplant organs. The use of AI models could improve clinicians’ decision-making capabilities and reduce diagnostic variability. In conclusion, our review highlights the advancements and limitations of AI in transplant pathology. We believe that these AI technologies have the potential to significantly improve transplant outcomes and pave the way for future advancements in this field.

Introduction. Recent trials demonstrated clinically significant benefits in HER2-nonamplified breast cancer with HER2-low expression using novel anti-HER2 antibody-drug conjugates. Thus, HER2-low breast cancer was proposed as a separate diagnostic entity. Herein, we reclassify HER2-negative cancers according to the new HER2-low category using a modified system and further investigate HER2-very-low expression. Methods. 114 HER2 immunohistochemistry (IHC)-negative invasive breast tumors were identified from the pathology database of Mayo Clinic, Jacksonville, FL, between January 2019 and August 2022. Two blinded breast pathologists (BP) independently rescored HER2 IHC slides at 200x and 400x magnification. Discordant cases between the two BPs were rescored together. The most recent 2018 ASCO/CAP HER2 scoring criteria were used. HER2 (0) was subdivided into HER2 (absent) and HER2 (very low). HER2 FISH testing was performed in all cases. Results. The cohort comprised of 38 (33.3%) HER2 (0) and 76 (66.7%) HER2 (1+) tumors. The first round of rescoring at 200x and 400x magnification resulted in 17 (14.9%) HER2 (absent), 31 (27.2%) HER2 (very low), and 64 (56.2%) HER2 (1+) and 2 (1.8%) HER2 (2+) tumors by BP1 and 20 (17.5%) HER2 (absent), 33 (28.9%) HER2 (very low), and 61 (53.5%) HER2 (1+) tumors by BP2. The combined final rescoring by BP1 and BP2 was as follows: 15 (13.2%) HER2 (absent), 35 (30.7%) HER2 (very low), 63 (55.3%) HER2 (1+), and 1 (0.9%) HER2 (2+) cases. A comparison of the first round of rescoring between two BPs showed substantial agreement with Cohen’s kappa value of 0.67. Both comparisons of first rescoring by BP1 and by BP2 to combined final rescoring showed almost perfect agreement with Cohen’s kappa value of 0.83.Follow-up FISH studies showed one amplified tumor. Conclusion. Our data support the need for finer granularity, classification, and understanding of HER2-low breast cancers. We also show that reproducibility between trained BP can be obtained, albeit with scoring at high power and low threshold for showing challenging interpretations.