A data-driven approach for chemotherapy recommendation model based on deep learning for patients with colorectal cancer in Korea

Background: Recently, the Clinical Decision Support System (CDSS) has attracted attention as a method for minimizing medical errors. To overcome the limitation that existing CDSS does not reflect actual data, we proposed CDSS based on deep learning. Methods: We proposed Colorectal Cancer Chemotherapy Recommender (C3R), a deep learning-based chemotherapy recommendation model. This supplements the limitation that the existing CDSS is difficult to support data-based decision making. It is configured to study the clinical data generated at Gachon Gil Medical Center and recommend appropriate chemotherapy. To validate the model, we compared the diagnosis concordance rate with the NCCN Guidelines, a representative cancer treatment guideline, and the results of the Gachon Gil Medical Center’s Colorectal Cancer Treatment Protocol (GCCTP). Results: The diagnosis concordance rates of the C3R model with the NCCN guidelines were 70.5% for the Top-1 Accuracy and 84% for the Top-2 Accuracy. Also, the diagnosis concordance rate with the GCCTP were 57.9% for the Top-1 Accuracy and 77.8 for the Top-2 Accuracy. Conclusions: This model is meaningful in that it is Korea’s first colon cancer treatment method decision support system that reflects actual data. In the future, if sufficient data is secured through multi-organization, more reliable results can be obtained.