AS-YOLO:A Novel YOLO model with Multi-scale Feature Fusion for Intracranial Aneurysm Recognition

Intracranial aneurysm is a common clinical disease that seriously endangers the health of patients. In view of the shortcomings of existing intracranial aneurysm recognition methods in dealing with complex aneurysm morphologies, varying sizes, as well as multi-scale feature extraction and lightweight deployment, this study proposes an intracranial aneurysm recognition method named AS-YOLO to improve the detection accuracy of intracranial aneurysms and adapt to the deployment on lightweight devices. This algorithm is based on the basic framework of YOLOv8n. Firstly, a cascaded fusion network is constructed to enhance the multi-scale feature extraction ability. Secondly, a multi level feature fusion module is designed to achieve more efficient multi-scale feature fusion. Then, the detection head is improved by proposing an efficient depthwise separable convolutional detection head , which reduces the number of model parameters and computational complexity while maintaining the detection accuracy. Finally, the SIoU loss function is introduced to make the bounding box regression of the model more accurate and further improve the detection performance. Compared with the original YOLOv8, the model proposed in this paper not only improves the detection accuracy in the aneurysm recognition task, with the mAP0.5 increased by 8.7% and the mAP0.5:0.95 increased by 4.96%, but also significantly reduces the number of model parameters and computational complexity, with the parameter quantity decreased by 8.21%. The model proposed in this paper combines multi-scale feature fusion and lightweight design, and can effectively meet the application requirements in resource-constrained environments such as mobile healthcare while ensuring high detection accuracy.