International Clinical and Medical Case Reports Journal (ISSN: 2832-5788) | Volume 4, Issue 2 | Original Article | Open Access DOI
Mostafa Daneshgar Rahbar*
Mostafa Daneshgar Rahbar*, Nabih Jaber, Daulet Kaldybek
Department of Electrical and Computer Engineering, Lawrence Technological University, Southfield, USA
*Correspondence to: Mostafa Daneshgar Rahbar
Fulltext PDFAim: This study aims to enhance intraoperative tissue characterization and classification in robotic bariatric surgery through a novel Dual Attention U-Net (DuAtUNet) model. By leveraging dual-channel attention mechanisms, DuAtUNet focuses on relevant features to achieve real-time, accurate segmentation of tissues, including fat, muscle, and vessels.
Methods: DuAtUNet was trained and tested using Python with OpenCV and Scikit-Image libraries on intraoperative images enhanced by vessel-specific filters. A dataset of surgical frames was used, with each image processed through dual channels to improve visibility of critical anatomical structures. Segmentation performance was evaluated by comparing U-Net, EUGNet, and DuAtUNet in terms of accuracy, Jaccard coefficient, and inference speed.
Results: DuAtUNet achieved higher accuracy and a better Jaccard coefficient compared to U-Net and EUGNet, demonstrating improved tissue differentiation and boundary precision. Specifically, DuAtUNet recorded an accuracy increase of 3% over the standard U-Net and a 4% improvement over EUGNet. Additionally, the Jaccard coefficient improved by 5% relative to U-Net and by 7% compared to EUGNet. The model’s attention mechanisms allowed for selective focus on critical regions, providing clearer segmentation and reducing background noise.
Conclusion: DuAtUNet significantly enhances intraoperative segmentation accuracy, supporting improved real-time visualization and decision-making in robotic-assisted bariatric surgery. This approach shows promise for broader applications in surgical environments requiring precise tissue characterization. Future studies should explore the integration of real-time deployment in clinical settings.
DuAtUNet; Attention mechanism; Robotic bariatric surgery; Tissue segmentation; Intraoperative classification; Vessel enhancement; Real-time AI
Mostafa Daneshgar Rahbar, Nabih Jaber, Daulet Kaldybek. Real-Time Intraoperative Tissue Characterization and Classification for Robotic Bariatric Surgery. Int Clinc Med Case Rep Jour. 2025;4(2):1-10.