University of Twente Student Theses
Automatic Detection of Cardiac Conduction Times on the ECG and Risk Prediction for High-Grade AV Block in TAVI Patients: Towards Feasible At-Home Monitoring
Velraeds, A.A.C. (2025) Automatic Detection of Cardiac Conduction Times on the ECG and Risk Prediction for High-Grade AV Block in TAVI Patients: Towards Feasible At-Home Monitoring.
|
PDF
19MB |
| Abstract: | Introduction: TAVI has become the main treatment for severe aortic stenosis in elderly patients. However, one of the most common complications is a risk of high-grade AV block due to destruction of the conduction system by the calcified valve. A high-grade AV block often requires a permanent pacemaker implantation. Predicting this complication based on conduction: times of the ECG could enable safe next day discharge and at-home monitoring. This thesis explores the feasibility of at-home ECG monitoring for TAVI patients, develops an algorithm to automatically detect conduction times, and constructs a prediction model for the risk of high-grade AV block. Methodology, Results & Discussion: Part I – The Feasibility of At-Home monitoring after TAVI: A study was conducted on the feasibility of at-home monitoring for TAVI patients receiving Kardia ECG devices by HartWacht in a cohort of 79 patients eligible for next day discharge. Although 68 patients submitted an ECG before TAVI, adherence to the full protocol (ECG recordings on day 1 and day 7 postTAVI) was low. Four patients followed the protocol correctly. The challenges included poor device use by only creating lead I ECGs and recalling issues. However, two cases of post-TAVI atrial fibrillation were successfully identified, demonstrating potential clinical value. Part II – The Automatic Detection of Cardiac Conduction Times: An algorithm was developed to automatically detect the heart rate, PQ-, QRS-, and QT intervals from hospital and Kardia ECGs. The hospital ECGs were used for development as there was a lack of Kardia ECGs and their reference values. Validation was performed using 20 manually annotated Kardia ECGs and references using the MUSE NX algorithm for the hospital ECGs. Parameter optimization was performed to obtain more accurate results. The Bland-Altman plots demonstrated that the algorithm is not yet accurate enough to be used in clinical practice. Part III – A Prediction Model for High-Grade AV Block after TAVI: Logistic regression models were trained using different (combinations of) conduction times from the preTAVI and postTAVI ECGs of TAVI patients. The most important predictors of high-grade AV block after TAVI included a wide preTAVI QRS interval and PQ interval prolongation comparing preTAVI and postTAVI. The final model achieved an AUC of 0.90, a sensitivity of 100%, and a specificity of 74% on the test set. These findings suggest that certain conduction times can serve as reliable predictors of high-grade AV block postTAVI. However, more data are needed before the implementation of this prediction model in clinical practice, due to the risk of overfitting in this dataset. General Discussion and Conclusion: This research demonstrates the technical and clinical potential of integrating automated cardiac conduction times analysis into at-home monitoring pathways for TAVI patients. The current feasibility for patients, by technical barriers, the accuracy of the algorithm, and the lack of data limit implementation in a clinical setting. Furthermore, the predictive model for high-grade AV block could assist clinicians identify high-risk patients. With improved patient instructions, reminder systems, interface design, and improved accuracy of the algorithm (e.g., by using a machine learning model) at-home ECG monitoring may become a safe and efficient standard for TAVI protocols, potentially allowing same-day discharge in the future. |
| Item Type: | Essay (Master) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 44 medicine, 50 technical science in general, 54 computer science |
| Programme: | Technical Medicine MSc (60033) |
| Link to this item: | https://purl.utwente.nl/essays/106469 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page