Predictive modelling for high-risk stage II colon cancer using auto-artificial intelligence.

Area of Science:

• Oncology
• Artificial Intelligence in Medicine
• Cancer Genomics

Background:

• Stratification of high-risk stage II colon cancer (CC) is crucial for determining the need for adjuvant chemotherapy.
• Previous methods for identifying high-risk CC patients have limitations.

Purpose of the Study:

• To define high-risk factors for recurrent stage II CC using artificial intelligence (AI).
• To develop a novel predictive model for identifying high-risk stage II CC patients.

Main Methods:

• A retrospective study of 259 stage II CC patients undergoing curative resection.
• Utilized Prediction One auto-AI software for predictive modeling and variable importance evaluation.
• Assessed disease-free survival (DFS) based on identified high-risk factors.

Main Results:

• The AI model achieved an area under the ROC curve (AUC) of 0.775.
• High-risk factors identified include elevated preoperative carcinoembryonic antigen (>5.0 ng/mL), venous invasion, and tumor obstruction.
• Patients with two or more high-risk factors exhibited significantly lower 5-year DFS (62.7%) compared to those with zero or one factor (87.4%).

Conclusions:

• A new predictive model using auto-AI software effectively identifies patients at high risk for stage II CC recurrence.
• Patients with two or more identified high-risk factors are candidates for adjuvant chemotherapy.