Prognostic significance of tumor budding in pancreatic carcinoma: Digitalized image approach evaluation using artificial intelligence. [version 1; peer review: awaiting peer review]

• Published in: F1000 research 2024, Vol.13, p.282
• Main Authors: Ben Rejeb, Sarra, Yaacoubi, Jasser
• Format: Article
• Language: English
• ISSN: 2046-1402; 2046-1402
• DOI: 10.12688/f1000research.146907.1

Introduction Pancreatic carcinoma (PC) is a highly malignant and lethal tumor characterized by a dismal prognosis which raised the need to identify other prognostic factors for better patient risk stratification. This study investigated the prognostic significance of tumor budding (TB) in pancreatic carcinoma using artificial intelligence. Methods In this retrospective multicenter study, we collected all cases of PC diagnosed (2008-2022). TB was assed using 2methods: manuel on hematoxylin-eosin (HE) slides and semi-automated using QUPATH software. The selected slide for each case was digitalized using NIS software version 4.00 connected to the microscope NIKON (Eclipse Ni-U). The pathological images were then incorporated into QUPATH. The budds were counted using cell count functionality based on the nucleus size and pixel variability, and TB scores were categorized as BUDD1(0-4), BUDD2(5-9) and BUDD3(≥10). We analyzed the association between the TB score and prognostic clinicopathological factors and overall survival. Results 25patients were included (mean age:62.3years;male-to-female ratio:2.57). TB was found in 100%of cases and a high TB score (BUDD2-3) was observed in 56%of cases (using QUPATH versus 48% using HE slides); statistical analysis showed no significant difference between the two methods(p=0.589). A high TB score was associated with older age(>72 years), ductal histological subtype and advanced stage (pT>2).53.8% of patients with lymph node metastasis or advanced stage had high TB score. Multivariate analysis revealed that TB score was strongly and independently associated with overall survival (OS), with a hazard ratio of 2.35. Conclusion TB is an additional prognostic factor in PC, and using artificial intelligence via QUPATH software offers a promising and accessible tool for pathologists to evaluate TB and to improve risk stratification in patients with PC.