A computer-aided diagnosis system is showing promise for predicting deeply invasive colorectal cancer but still lags behind established methods.

• The CAD system achieved an overall diagnostic accuracy of 61.3%, with high confidence cases showing 95.3% accuracy.
• Compared to magnifying narrow-band imaging (m-NBI) and magnifying chromoendoscopy (mCE), the CAD system’s performance was inferior (p = 0.011 and p = 0.014, respectively).

For surgeons, integrating high-confidence CAD data may enhance decision-making, but reliance on traditional methods remains crucial for low-confidence cases.

• m-NBI and mCE demonstrated superior sensitivity (74.2% and 83.9% respectively) and were more reliable overall.