YOLOv8-IE

Neuroblastoma exhibits diverse clinical subtypes, making early and accurate diagnosis challenging despite the effectiveness of CT imaging for tumor detection. To address subtype similarity, this study proposes a modified YOLO-based model, YOLOv8-IE, which integrates inverse residual attention (iRMB) and a centered feature pyramid (EVC) to enhance feature focus and fusion. The proposed YOLO-IE model achieved a 7.9% improvement in mean Average Precision over the baseline YOLO, demonstrating the strong potential of AI-driven methods for neuroblastoma detection and classification.