Supervisor: Dr. Rory Turnbull | SUSTAIN Facility | University of Leeds
Background:
Equipment and consumable use in surgical procedures is often inefficient and wasteful. Unused tools are routinely re-sterilized at significant economic and environmental cost, contributing to a substantial healthcare carbon footprint. Automating "tray rationalisation" through computer vision—where AI models are trained to identify and track surgical instruments—offers a pathway to streamline inventory and minimize waste.
While deep learning models have demonstrated high potential in identifying carefully arranged medical instruments, real-world operating rooms are highly dynamic environments. During a procedure, instruments frequently become disorganized, overlap, and are obscured by bodily fluids or harsh lighting. This project explores the operational threshold of object detection AI under varying degrees of clinical pressure. The goal is to investigate how digital asset tracking can be integrated into high-stress environments to achieve Net Zero targets, without necessitating physical changes to standard trays or adding psychological burden to the surgical team.
Personal motivation:
As a Chemical Engineering student interested in transitioning into Biomedical Engineering, I am deeply passionate about advancing sustainability within healthcare systems. Growing up in Algeria, I observed that healthcare infrastructure and environmental sustainability are often treated as entirely separate challenges, and both are frequently under-resourced.
This project presents a unique opportunity to bridge that gap. I look forward to developing new technical capabilities in computer vision and digital healthcare, while directly applying the systems-level process optimization principles I have acquired during my engineering studies to build more resilient, sustainable clinical pathways.
Methodology:
I will conduct a comparative data analysis utilizing the high-fidelity simulated surgery environment at the SUSTAIN facility. I will build a multi-condition image dataset of standard surgical trays, split into two testing scenarios:
-
Baseline Testing: Capturing clean, perfectly arranged trays prior to surgical use.
-
Stress Testing: Photographing the same trays under simulated clinical disorder, introducing heavy instrument occlusion, different lightings, and simulated bodily fluids.
The surgical instruments will be provided by the School of Dentistry at the University of Leeds.
After annotating the dataset, I will implement a YOLO-based framework (AI model) to process the images and statistically map where detection performance fluctuates.
Outputs and Impact:
By establishing a framework for AI-driven tray rationalisation, this research targets the elimination of unnecessary sterilization cycles. Successfully tracking unused assets will yield significant economic savings, lower the hospital's carbon footprint, and support the barrierless implementation of sustainable healthcare technology.