Evaluation of Artificial Intelligence Algorithms Performance in the Histopathological Diagnosis of Basal Cell Carcinoma

Background Basal cell carcinoma (BCC) is the most commonly diagnosed skin cancer worldwide and the predominant form of non-melanoma skin cancers (NMSCs), with an escalating global incidence. Histopathology remains the gold standard for diagnosis; however, manual analysis is labor-intensive, time-consuming, and subject to increasing pressure amid a global shortage of board-certified dermatopathologists. Digital pathology and whole-slide imaging (WSI), combined with advanced artificial intelligence (AI) models such as vision transformers and large language models (e.g., HistoGPT), offer a transformative solution to automate and streamline dermatopathological diagnostics.

Aim of the Work This study aims to evaluate the diagnostic performance and processing efficiency of the AI algorithm HistoGPT in the histopathological diagnosis of basal cell carcinoma compared to certified dermatopathologists.

Methodology This retrospective, blinded, comparative study will be conducted using archived H&E-stained glass slides retrieved from the pathology archive of the Al-Hussein Dermatopathology Unit between 2010 and 2019. Slides meeting the inclusion criteria will be digitized into high-resolution Whole Slide Images (WSIs) at 40× magnification using the Leica Aperio GT450 scanner. The digitized WSIs will be processed and analyzed through the HistoGPT cloud platform to automatically generate diagnostic reports and classifications. The AI-generated findings will be systematically compared against the reference standard diagnoses established by a panel of certified dermatopathologists under the supervision of Prof. Hussein Hasb El-Nabi. Diagnostic accuracy, concordance, and turnaround time will be evaluated.

Statistical Analysis Data will be analyzed using SPSS (version 26.0) or R-programming. Categorical variables will be compared using appropriate statistical tests, and a $p$-value of $< 0.05$ will be considered statistically significant.