Deep learning could predict leakage during gastric cancer surgery. High-resolution imaging, single-image analysis, and data augmentation were pivotal for model performance. These findings lay the groundwork for clinical applications and future research on surgical image analysis.