An auxiliary detection model for gastrointestinal nursing based on deep learning

Gastrointestinal polyps are early pathological lesions of the digestive tract and major precursors to colorectal cancer. Early and accurate detection is, therefore, critical for guiding timely clinical nursing interventions and ultimately reducing cancer incidence. In routine endoscopy and nursing assessment, however, polyp detection remains challenging: their subtle and concealed morphology, illumination-related artifacts, and clinician visual fatigue frequently lead to missed lesions and false alarms. To mitigate these issues, we propose a lightweight yet high-accuracy intelligent nursing-assisted detection framework based on an enhanced YOLOv8 architecture. In particular, we introduce a PPAC2f module into the backbone to reinforce salient-region perception and fine-grained feature modeling. In addition, we design a CGAFusion module to facilitate effective cross-layer interaction and aggregation of spatial and semantic representations. Together, these innovations improve robustness under complex endoscopic conditions. Experiments on the Kvasir-SEG dataset and selected wireless capsule endoscopy samples demonstrate that our method achieves state-of-the-art performance, consistently outperforming YOLOv8n, YOLOv10n, YOLOv11n, and RT-DETR in Precision, Recall, mAP@0.5, and mAP@0.5-0.95. Notably, the proposed model attains 94.7% mAP@0.5 and 70.3% mAP@0.5-0.95, substantially reducing both missed detections and false positives. These results suggest that the model provides reliable AI-assisted support for gastrointestinal endoscopy and exhibits strong potential for integration into clinical nursing practice.