202. AN ARTIFICIAL INTELLIGENCE ALGORITHM FOR DETECTION OF CIRCULATING TUMOR CELLS IN PATIENTS WITH ESOPHAGEAL CANCER

Abstract Background Circulating tumor cells (CTCs) are considered predictors of distant metastasis, therapeutic response, and prognosis in various cancer. However, the overall process of detecting these marker expressions in CTCs becomes more complex and time consuming as more markers of cytopathologic heterogeneity are discovered. The aim of this study was to evaluate the use of a convolutional neural network (CNN)-based image processing artificial intelligence (AI) for CTC detection in ESCC patients. Methods We first investigated the accuracy of the AI algorithm to distinguish ESCC lines (KYSEs) from peripheral blood mononuclear cells (PBMCs). Then we used the AI algorithm to detect CTCs in peripheral blood samples obtained from ESCC patients. Results The AI algorithm discriminated KYSE from PBMCs of healthy volunteers accompanied by EpCAM staining with greater than 99.8% accuracy when the AI was trained on the same KYSE. The average accuracy of the AI and the 4 investigators in the differentiation of KYSE from PBMCs was 100% and 91.8%, respectively (p = 0.011). Average time to complete classification of 100 cell images for AI and human was 0.74 and 630.4 seconds, respectively (p = 0.012). The average number of EpCAM-positive/DAPI-positive cells detected in blood samples by the AI were 44.5 over 10 ESCC patients and 2.4 over 5 healthy volunteers, respectively (p = 0.019). Conclusion These results indicate that the CNN-based image-processing algorithm for CTC detection provides higher accuracy and shorter analysis time compared to human, suggesting its clinical application for ESCC patients.