Evaluating the performance of LYDIA: an AI-powered decision support system in detection of metastatic tumors in lymph nodes

The integration of AI in histopathology represents a significant advancement in diagnosing metastatic cancers. LYDIA (LYmph noDe assIstAnt) is a commercially available AI-powered tool designed to annotate tumors in lymph node sections to aid histopathologists in diagnosing metastases. This study rigorously evaluates LYDIAs standalone performance and the time and cost benefits on diagnostic workflow. In this study, LYDIAs performance was rigorously evaluated on a blind image dataset comprising 366 whole slide images (WSIs) from metastatic breast, colon, lung, and skin cancers. Additionally, a clinical diagnostic workflow was simulated using an internal cohort of 105 WSIs, evaluated by four experienced histopathologists, to assess the impact of LYDIA when used as a decision support tool on diagnostic accuracy and case handling time. A Monte Carlo simulation was also conducted to estimate potential reductions in Immunohistochemistry (IHC) costs associated with AI assistance in sentinel lymph node evaluation. The analysis demonstrated that LYDIA achieved excellent diagnostic accuracy with high ROC-AUC scores (0.995 for breast cancer, 0.963 for colon cancer, 0.973 for lung cancer, and 0.983 for melanoma). In a clinical setting, AI-assisted diagnosis significantly reduced time-to-diagnosis across all metastasis sizes, with a maximum of 1.59-fold acceleration for micro-metastases (saving 26.5 seconds per WSI). LYDIA also enhanced pathologists diagnostic performance, increasing their sensitivity from 77.3% to 87.3%. The Monte Carlo simulation indicated an average saving of {euro}7.89 per case, highlighting cost benefits from reduced IHC requests. In conclusion, the findings highlight LYDIAs capabilities as a supportive tool that holds the potential to reduce diagnostic turnaround times and assist pathologists in identifying important regions within slides. This study illustrates the capability of AI-powered solutions like LYDIA to accelerate diagnosis and ultimately improve patient care. The observed time and cost benefits suggest that integrating AI into routine pathology practice can improve diagnostic quality, reduce costs and optimize resource utilization.