Endoscopy event recognition using machine learning

点击翻译按钮获取中文摘要

1) Brief description of the context of the research including potential impactAn endoscopic procedure such as colonoscopy is defined by a series of events these include reaching thecaecum, withdrawal, detecting/diagnosing disease, resection and removal. The quality of a procedure ismeasured by metrics related to these events e.g. % of procedures where caecum is reached, number ofpolyps detected etc. There are multiple studies that show the quality of colonoscopy varies significantlyamongst doctors and that the recording and identification of these events is inconsistent.The scope of this project is to explore machine learning to improve the detection and documentation of events in colonoscopy with the goal of improving the quality of procedures. 2) Aims and ObjectivesThe specific objectives are to:Develop video-based methods that include temporal information for event detection. 3) Novelty of Research MethodologyThis is a highly challenging topic that will address open research questions related to temporal machine learning with small amounts of datadomain adaption for data generalization and multi-task learning that can use data from other endoscopic events to improve performance.4) Alignment to EPSRC's strategies and research areasThis project lies within the healthcare technology theme helping to accelerate translation within the scope of the following grand challenges:Developing future therapiesFrontiers of physical interventionOptimising treatmentTransforming community health and care5) Any companies or collaborators involvedOdin Medical Limited