Comprehensive Summary
Cha et al. used deep learning based algorithms to determine if accurate prediction of lethal prostate cancer (PCa) metastasis after radical prostatectomy risk can occur. Previous models were often multimodal, utilizing both pathological and clinical data. However, the researchers created a model to compare the performance of their histopathology diagnostic tool against existing standard clinical risk tools and genomic classifiers. Utilizing representative whole slide images (WSIs) from RP and needle biopsy samples, or tissue microarrays (TMAs) constructed from RP specimens, they developed a classification system based on this data. The model was found to accurately predict metastasis with similar prognostic performance in comparison to genomic classifiers and existing clinical data. In conclusion, Cha et al. developed this model with sole reliance on tumor tissue histopathology to accurately predict lethal PCa metastasis that performs comparably to its expensive and time-consuming counterpart (genomic classifiers), enabling future diagnostic and therapeutic development.
Outcomes and Implications
This study and model developed provides an alternative diagnostic tool in accurate metastasis prediction that is more accessible, cost-effective, and time-saving, which can allow for possible higher level stratification. In addition, PCa treatment is highly dependent on accurate assessment of the tumor's behavior and aggressiveness; an accurate tool can allow for better diagnostics and treatment plans to create more targeted personalized treatments while also helping certain patients avoid unnecessary treatments. This tool can create a more standardized and reliable prognoses for patients by quantifying tumor morphology (reduced inter-observer variability).