Enhanced YOLOv8n-seg is used for the research on spinal segment segmentation based on machine vision within the surgical visual field

View at publisher Summarise with VIRAT
Authors: Yuheng Sun, Yun Dong, Xiaopan Cai
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Intraoperative spinal segmentation remains challenging due to uneven illumination, tissue occlusion, and complex anatomical textures that hinder accurate boundary perception. To address these issues, we propose an enhanced lightweight YOLOv8n-seg model integrating attention and boundary refinement mechanisms. The attention module adaptively emphasizes salient spatial and channel features, while the boundary refinement head improves contour smoothness and continuity, particularly for slender and low-contrast spinal structures. A dataset of approximately 1,000 annotated intraoperative spinal images was established through active learning and structured augmentation. Experimental results demonstrate that the proposed model achieves an mAP@0.5 of 0.927 and mAP@0.5–0.95 of 0.534 at 110 FPS, maintaining real-time performance while improving segmentation precision.

Fields of Study & Subjects

Medical Imaging and Analysis Advanced Neural Network Applications Medical Image Segmentation Techniques

Publication details

Journal / source
International Conference on Imaging Processing for Medical-Oriented Recognition (IPOR)
Pages
16–16
Publication date
19 March 2026
Language
EN