A Deep Learning Framework for Assessing the Risk of Transvenous Lead Extraction Procedures

Wahid, Fazli, Ma, YingLiang ORCID: https://orcid.org/0000-0001-5770-5843, Mehta, Vishal, Howell, Sandra, Niederer, Steven and Rinaldi, C. Aldo (2024) A Deep Learning Framework for Assessing the Risk of Transvenous Lead Extraction Procedures. In: Artificial Intelligence in Healthcare - 1st International Conference, AIiH 2024, Proceedings. Lecture Notes in Computer Science, 14976 . Springer, GBR, pp. 17-30. ISBN 9783031672842

PDF (DeepLearningForTLE_AIHealth_final) - Accepted Version Restricted to Repository staff only until 15 August 2025. Request a copy

Abstract

This paper introduces a deep-learning framework augmented with human guidance for evaluating the risk associated with Transvenous Lead Extraction (TLE). TLE is one type of minimally invasive cardiac procedures, and it is to remove old pacing wires inside the heart. The deep-learning framework automatically extracts geometric features from a single plain chest X-ray image obtained before the procedure. It then utilizes these features in conjunction with clinical data to predict the procedural risk. All geometric features were recommended by a senior clinician and include the positions of coils, the number of leads inside the superior vena cava and the angle of leads. The proposed framework was trained and tested using a database comprising records from 1,053 patients who underwent TLE procedures. Notably, the framework was successfully trained despite the highly imbalanced nature of the data. An accuracy of 0.91 was achieved and the framework can predict 88% of major complication cases. By combining geometric features with clinical data, we were able to deliver a significantly better accuracy and a higher recall rate for detecting high-risks patients, when compared with existing approaches. The methodology described in this paper can be applied to the risk assessment for other cardiac procedures.

Item Type:	Book Section
Uncontrolled Keywords:	deep learning,geometric feature extraction,risk assessment,theoretical computer science,computer science(all) ,/dk/atira/pure/subjectarea/asjc/2600/2614
Faculty \ School:	Faculty of Science > School of Computing Sciences
UEA Research Groups:	Faculty of Science > Research Groups > Norwich Epidemiology Centre Faculty of Medicine and Health Sciences > Research Groups > Norwich Epidemiology Centre
Related URLs:	http://www.scopus.com/inward/record.url?...
Depositing User:	LivePure Connector
Date Deposited:	20 Aug 2024 11:30
Last Modified:	24 Sep 2024 08:28
URI:	https://ueaeprints.uea.ac.uk/id/eprint/96271
DOI:	10.1007/978-3-031-67285-9_2

Actions (login required)

View Item