Evaluating AI-Based Mitosis Detection for Breast Carcinoma in Digital Pathology: A Clinical Study on Routine Practice Integration

Background/Objectives: An accurate assessment of mitotic activity is crucial in the histopathological diagnosis of invasive breast carcinoma. However, this task is time-consuming and labor-intensive, and suffers from high variability between pathologists. Methods: To assist pathologists in routine diagnostics, we developed an artificial intelligence (AI)-based tool that uses whole slide images (WSIs) to detect mitoses, identify mitotic hotspots, and assign mitotic scores according to the Elston and Ellis grading system. To our knowledge, this study is the first to evaluate such a tool fully integrated into the pathologist’s routine workflow. Results: A clinical study evaluating the tool’s performance on routine data clearly demonstrated the value of this approach. With AI assistance, pathologists achieved a greater accuracy and reproducibility in mitotic scoring, mainly because the tool automatically and consistently identified hotspots. Inter-observer reproducibility improved significantly: Cohen’s kappa coefficients increased from 0.378 and 0.457 (low agreement) without AI to 0.629 and 0.726 (moderate agreement) with AI. Conclusions: This preliminary clinical study demonstrates, for the first time in a routine diagnostic setting, that AI can reliably identify mitotic hotspots and enhance pathologists’ performance in scoring mitotic activity on breast cancer WSIs.