OC-065 PEEKABOO WITH MESHES AND MACHINES: DEEP LEARNING TO IDENTIFY TACKS AND MESHES ON CT SCANS. (13th October 2022)

• Record Type: Journal Article
• Authors: Rengan, V; Meenashi Sundaram, P; Arora, E; Bawa, A
• Is Part Of: British journal of surgery. Volume 109(2022)Supplement 7
• Publication Date: 2022-10-13
• Subjects: Surgery -- Periodicals; 617.005
• DOI: 10.1093/bjs/znac308.077
• Languages: English
• ISSNs: 0007-1323

Abstract

Abstract: Aim: To identify the presence of tacks and meshes in CT scans of patients who have previously undergone hernia repairs Materials & Methods: We annotated data from >100 anonymised Hernia CT scans of patients who had undergone previous ventral hernia repairs and used machine learning(ML)/ deep learning(DL) techniques to identify the presence of meshes and tacks.Annotation of CT scans was performed using computer vision tools. A combination of image processing, feature extraction and artificial intelligence(AI) techniques were used to create models that could identify tacks and meshes on CT scans. Results: We were able to identify the presence of meshes and tacks with >55% accuracy. The AI model under construction is continuing to improve performance. Conclusions: Identification of the mesh is a challenge for surgeons and radiologists alike. The presence of machine learning techniques has revolutionised radiology and made the identification of obscure structures possible. Meshes present a unique challenge as they constitute a foreign tissue which integrates into native tissue. We believe that identification of hernia meshes and tacks can help surgeons identify the right planes for planning re-surgeries while operating in patients with previous hernia repairs or while tackling recurrences.