Combining a deep learning model with clinical data better predicts hepatocellular carcinoma behavior following surgery.

Hepatocellular carcinoma (HCC) is among the most common cancers worldwide, and tumor recurrence following liver resection or transplantation is one of the highest contributors to mortality in HCC patients after surgery. Using artificial intelligence (AI), we developed an interdisciplinary model to predict HCC recurrence and patient survival following surgery. We collected whole-slide H&E images, clinical variables, and follow-up data from 300 patients with HCC who underwent transplant and 169 patients who underwent resection at the Cleveland Clinic. A deep learning model was trained to predict recurrence-free survival (RFS) and disease-specific survival (DSS) from the H&E-stained slides. Repeated cross-validation splits were used to compute robust C-index estimates, and the results were compared to those obtained by fitting a Cox proportional hazard model using only clinical variables. While the deep learning model alone was predictive of recurrence and survival among patients in both cohorts, integrating the clinical and histologic models significantly increased the C-index in each cohort. In every subgroup analyzed, we found that a combined clinical and deep learning model better predicted post-surgical outcome in HCC patients compared to either approach independently.

The significance of focally enhanced gastritis in haematopoietic stem cell transplant recipients.

AIMS: The histological diagnosis of acute gastric graft-versus-host-disease (aGVHD) in patients with a history of haematopoietic stem cell transplant (HSCT) is based on the presence of epithelial cell apoptosis and karyorrhectic debris. There is, however, limited information on the histological findings in patients who develop symptoms several months after transplant. Focally enhanced gastritis (FEG), defined by the presence of focal periglandular lymphohistiocytic inflammation with neutrophilic or lymphocytic intra-epithelial infiltration of gastric glands, has been described in patients with inflammatory bowel disease and in HSCT patients. The pattern closely resembles the focal periductal inflammation and lymphocytic exocytosis seen in chronic GVHD of the salivary gland. We sought to evaluate the significance of FEG in HSCT patients. METHODS AND RESULTS: Gastric biopsies from 151 HSCT patients who underwent endoscopies for GVHD-like symptoms were identified. Time from transplant to biopsy, presence of extra-gastric GVHD, medications and outcome were noted. Thirty-five biopsies showed FEG and 21 showed aGVHD; the remainder were either normal or showed non-specific changes. Twenty-one (60%) FEG patients had concurrent histologically proven extra-gastric GVHD. The time to biopsy in FEG patients was significantly longer than in aGVHD patients (162 versus 57 days, P < 0.01). Prior or subsequent gastric biopsies of 14 patients in the FEG cohort were also evaluated and, of these, six showed aGVHD while one showed persistent FEG. CONCLUSIONS: These findings suggest that FEG probably represents a form of late-occurring GVHD. This histological pattern should not be overlooked when identified in gastric biopsies from HSCT patients.

High-grade urothelial carcinoma in urine cytology: different spaces - different faces, highlighting morphologic variance.

INTRODUCTION: The Paris System for Reporting Urinary Cytology (TPS) was first published in 2016 to standardize reporting and placed a specific emphasis on high-grade urothelial carcinoma (HGUC). The urinary tract is anatomically divided into the upper tract (UT) and the lower tract (LT). A major morphologic criterion in TPS for HGUC defines the nuclear-to-cytoplasmic (N/C) ratio as ≥ 0.7. In this study, we evaluated N/C ratios of HGUC arising from UT and LT urine specimens, to ascertain differences due to location. MATERIALS AND METHODS: Digital annotations of whole slide scanned images were performed and enumerated. RESULTS: The cohort consisted of 59 ThinPrep specimens from 52 patients. The majority of the tumors were located in LT (39 of 59, 66.1%). A total of 590 cells were analyzed (10 cells per case). In UT, the average N/C was 0.58 and LT the average was 0.54 (P < 0.001). The average nuclear area for UT was 126.3 and for LT was 158.2 µm2 (P = 0.01). The average cytoplasmic area for UT was 219.1 µm2 and for LT was 296.2 µm2 (P < 0.001). The average cellular circumference for UT was 59.4 µm and for LT was 66.1 µm (P < 0.001). CONCLUSIONS: We found that UT HGUCs have higher N/C ratios, smaller cell circumference, smaller nuclei, and less cytoplasm compared with LT. When UT was divided into renal pelvis and ureter, no statistical difference was identified.

Digital image analysis supports a nuclear-to-cytoplasmic ratio cutoff value below 0.7 for positive for high-grade urothelial carcinoma and suspicious for high-grade urothelial carcinoma in urine cytology specimens.

BACKGROUND: Urinary cytology is sensitive and specific for diagnosing and screening high-grade urothelial carcinomas (HGUC). The Paris System (TPS) for urinary cytology was introduced in 2016 to standardize reporting. According to TPS diagnostic categories of HGUC and suspicious for HGUC (SHGUC), the average nuclear-to-cytoplasm (N:C) ratio of atypical cells should be ≥0.7. The objective of the current study was to measure the N:C ratio of urine cytology specimens with HGUC and SHGUC diagnoses and biopsy-proven HGUC follow-up. METHODS: A cohort of 64 cases (HGUC, 49 cases; SHGUC, 15 cases) from 57 patients was constructed. Urine cytology slides were scanned into whole-slide digital images. The nuclear and cytoplasmic areas were enumerated by digital image analysis (DIA), and the N:C ratios were measured. RESULTS: In total, 640 cells were analyzed by DIA (HGUC, 490 cells; SHGUC, 150 cells). For HGUC and SHGUC, the average N:C ratios were 0.57 and 0.53, respectively. The maximum average N:C ratio was 0.73 for HGUC and 0.68 for SHGUC. HGUC had higher average N:C ratio (P < .001), higher average nuclear area (P < .001), higher average maximum N:C ratio (P = .005), and higher average maximum nuclear area (P = .006) compared with SHGUC. CONCLUSIONS: The N:C ratios for the HGUC (0.57) and SHGUC (0.53) categories are lower than those previously suggested in TPS. The authors advocate reducing the N:C ratio below the current threshold of 0.7.