Innovative strategy for colorectal cancer with artificial intelligence

Introduction: Although surgical resection with lymph node dissection is a standard treatment strategy for T2 colorectal cancer (CRC), 25% of patients are lymph node metastasis (LNM)-positive. The remaining LNM-negative patients may theoretically choose endoscopic full-thickness resection (EFTR) or transanal endoscopic microsurgery (TEM) alone if we can predict LNM after EFTR or TEM. This study investigated whether artificial intelligence (AI) can predict the presence of LNM and proposed “Resect and Analysis” strategy for T2 CRC, as achieved for T1 CRC.
Methods: In total, 511 consecutive patients with T2 CRCs surgically resected were included. We divided patients into two groups: 411 patients were used for machine learning for the AI model, and the remaining 100 patients were used for model validation. The AI model analyzed seven factors (patient age, sex, tumor size, location, tumor differentiation, lymphatic invasion, vascular invasion), and an AI model was developed to predict the likelihood of LNM occurrence. The AI model was validated by calculating the area under the receiver operator characteristic curve (AUC) for predicting LNM.
Results: The rates of LNM-positive in the training and validation datasets were 26 (106/411) and 28% (28/100), respectively. The AUC of the AI model was 0.96. The sensitivity and specificity of the AI model were 94 and 96%, respectively.
Conclusion: AI can accurately predict the presence of LNM in patients with T2 CRC, leading “Resect and Analysis” potential strategy.