Machine Learning Algorithms and AI: Augmenting Safety and Predicting Complications in Transcatheter Aortic Valve Replacement

Authors: Zain Khalpey, MD, PhD, FACS, Kirtana Roopan, Ujjawal Kumar BA, Jessa Deckwa BS, Robert Burke, MD, Robert Riley, MD, and David Rizik, MD

Abstract:

This article explores the potential of machine learning (ML) algorithms and artificial intelligence (AI) to improve the safety and predict complications of transcatheter aortic valve replacement (TAVR). We examine the impact of ML and AI on patient selection, procedure planning, intraprocedural guidance, and post-procedural monitoring. The potential benefits of these technologies include increased efficiency, reduced complications, and improved patient outcomes. We also discuss the challenges and future directions in the development of AI-powered TAVR.

1. Introduction

Transcatheter aortic valve replacement (TAVR) has emerged as a minimally invasive alternative to surgical aortic valve replacement for patients with severe aortic stenosis. The procedure involves the implantation of a prosthetic valve through a catheter, avoiding the need for open-heart surgery. Although TAVR has shown promising results in terms of reduced morbidity and mortality, complications can still occur. Machine learning algorithms and artificial intelligence have the potential to transform the way TAVR procedures are performed, ultimately improving patient safety and predicting complications.

2. Machine Learning and AI in Patient Selection for TAVR

TAVR, although minimally invasive, is not without its risks and complications. Risk stratification tools and models have steadily accrued to better inform cardiologists about what risk factors and most important to consider when approaching TAVR. In a retrospective study of the 52,000 patients from the Society of Thoracic Surgeons (STS) Registry, it was shown that one-year mortality post-TAVR was strongly associated previous stroke, major, life-threatening or disabling bleeding, stage III acute kidney injury, and moderate or severe perivalvular regurgitation.1 It was also shown that surgery centers that were novice to the TAVR procedure, or had not performed many, were more likely to have complications.

Machine Learning could help improve TAVR by attacking multiple aspects of quality improvement elucidated in this retrospective study. Algorithms can be designed to improve preoperative risk stratification and intraoperative guidance. Post-procedural monitoring would also be improved with AI-driven tools, helping predict and identify complications. Artificial Intelligence is already improving TAVR, but here is how we view it continuing to increase precision with TAVR procedures.

2.1. Risk Assessment and Stratification

ML algorithms can analyze large amounts of patient data to identify pre-procedural risk factors for complications, thereby aiding in the selection of appropriate candidates for TAVR. Evertz et al trained an AI-powered function to analyze ejection fraction and other echocardiographic evidence to pre-operatively stratify patients for TAVR, showing equivalent stratification to manual assessment of heart function (2). Further AI-driven tools such as this one can stratify patients based on their preoperative risk profiles, enabling clinicians to make more informed decisions about treatment options.

2.2. Imaging Analysis and Interpretation

ML algorithms can analyze pre-procedural imaging data (e.g., computed tomography, echocardiography) to predict the optimal valve size and type for each patient, as seen in Willson et al (3). AI-powered software coupled with new-technology can identify comorbidities during preoperative evaluation of heart function as seen in Gohmann et al. In this study, CT-FFR was used to evaluated TAVR patients for co-morbid CAD with reasonable success (4).

3. Procedure Planning and Intraprocedural Guidance

3.1. Procedural Simulation and Optimization

Automated and Semi-automated approaches were compared to standard of care analysis of aortic annular measurements for aortic stenosis in 3D TEE image. Semi-automated analysis using the novel automated software reduced analytical time while maintaining similar accuracy compared to the conventional cross-sectional analysis (5). AI-driven simulations can optimize the TAVR procedure by identifying the best approach, valve size, and implantation technique.

Another study sought to develop and compare an array of machine learning methods to predict in-hospital mortality after transcatheter aortic valve replacement (TAVR) in the United States. By retrospectively reviewing a total of 10,833 TAVRs they were able to evaluate the morbidity and mortality. The best model was obtained by logistic regression (area under the curve: 0.92; 95% confidence interval: 0.89 to 0.95) (6). Most obtained models plateaued after introducing 10 variables. Acute kidney injury was the main predictor of in-hospital mortality ranked with the highest mean importance in all the models. ML algorithms can analyze previous procedural data to predict potential complications and suggest strategies to mitigate them (6).

3.2. Intraprocedural Image Analysis and Guidance

ML algorithms can identify potential complications (e.g., valve malposition, coronary obstruction) and provide guidance on how to address them (8). Current prognostic risk scores for transcatheter aortic valve implantation (TAVI) do not benefit yet from modern machine learning techniques, which can improve risk stratification of one-year mortality of patients before TAVI. A validation study was performed in two centers and evaluated 1300 and 631 patients who underwent TAVI. Models developed with the larger dataset offered similar or higher prediction accuracy on the external validation (8). Logistic regression, random forest and CatBoost lead to areas under the curve of the ROC of 0.65, 0.67 and 0.65 for the internal validation and of 0.62, 0.66, 0.68 for the external validation, respectively (8).

As machine learning models continue to integrate into medicine the application for different data inputs can be explored. For example, a study was performed to evaluate the doppler velocity index (DVI) following a TAVR procedure which could indicate risk of complications. Patients with severe aortic stenosis enrolled in the PARTNER (Placement of Aortic Transcatheter Valve) 2 (intermediate surgical risk) or PARTNER 3 (low surgical risk) trial undergoing TAVR (n = 1,450) or SAVR (n = 1,303) were included (8). Following TAVR, there were no differences among the 3 DVI groups in composite outcomes of death, stroke, or rehospitalization or in any individual components of 2-year outcomes (P > 0.70 for all). Following SAVR, there was no difference among DVI groups in the composite outcome (P = 0.27), but there was a significant association with rehospitalization (P = 0.02) (8). Although no significant changes based on the findings were suggested using a restricted cubic-spline analysis, the implementation of machine learning could help us identify which parameters could lead to increased DVI post-operatively and then further suggest which patients should undergo a TAVR instead of a surgical valve repair/replacement.

This is just the beginning of analysis of imaging values pre- and post-TAVR. By expanding into the AI into the medical realm, we could accelerate the ways in which treatment and/or incite for continued patient monitoring for a specific complication following a TAVR procedure.

4. Post-procedural Monitoring and Complication Prediction

4.1. Remote Monitoring and Predictive Analytics

In one study, a wearable sensor system is used to continuously transmit ECG and contextual data to a central monitoring unit, allowing remote follow-up of ECG abnormalities and physical deteriorations. Telemonitoring is suggested as an alternative or supplement to current in-hospital monitoring after TAVR, enabling early hospital dismissal in eligible patients and accessible follow-up prolongation. (7)

Another study evaluated one year morbidity and mortality prediction with external validation of the algorithm developed using a different center. The AI-driven remote monitoring systems can track patient data (e.g., vital signs, valve function) and alert clinicians to potential complications (9). Machine learning by extreme gradient boosting was trained and tested with repeated 10-fold hold-out testing using 34 pre- and 25 periprocedural clinical variables. In external validation the machine learning model had an area under the receiver-operator-curve of 0.82 (0.78-0.87) for prediction of 1-year all-cause mortality following hospital discharge after TAVI which was superior to pre- and periprocedural clinical variables including age 0.52 (0.46-0.59) and the EuroSCORE II 0.57 (0.51-0.64), p < 0.001 for a difference (9).

A similar study was performed using gradient boosting machine learning (GBM) prediction model included 163 variables and was optimized using 5-fold cross-validation repeated 10-times (10). The receiver operator characteristic (ROC) for the GBM model was calculated to predict one-year mortality post TAVR, and then compared to the TAVI2-SCORE and CoreValve score. Based on GBM, the ten most predictive variables for one-year survival were cardiac power index, hemoglobin, systolic blood pressure, INR, diastolic blood pressure, body mass index, valve calcium score, serum creatinine, aortic annulus area, and albumin (10). The area under ROC to predict survival for the GBM model vs TAVI2-SCORE and CoreValve Score was 0.72 (95% CI 0.68-0.78) vs 0.56 (95%CI 0.51-0.62) and 0.53 (95% CI 0.47-0.59) respectively with p < 0.0001 (10). ML Algorithms can give insight on which pre-operative parameters can contribute to postoperative complications and presence of these parameters could affect the approach a surgeon suggests.

4.2. Personalized Patient Care and Follow-up

AI-powered tools can tailor follow-up care plans based on individual patient characteristics and predicted risk of complications. The use of machine learning in complicated care practices will require ongoing consideration, since the correct diagnosis in a particular case and what constitutes best practice can be controversial (11). Prematurely incorporating a particular diagnosis or practice approach into an algorithm may imply a legitimacy that is unsubstantiated by data (11).

5. Challenges and Future Directions

5.1. Data Quality and Standardization

The collective medical mind is becoming the combination of published literature and the data captured in health care systems, as opposed to individual clinical experience (11). The shift is presenting new and exciting ways to incorporate learning from aggregated data with machine learning and artificial intelligence. If clinicians remain ignorant about the construction of machine-learning systems or allowing them to be constructed as black boxes could lead to ethically problematic outcomes (11).

Ensuring the quality and standardization of data used to train ML algorithms is essential for the development of accurate and reliable AI-driven tools. Algorithms introduced in non-medical fields have already been shown to make problematic decisions that reflect biases inherent in the data used to train them. For example, programs designed to aid judges in sentencing by predicting an offender’s risk of recidivism have shown an unnerving propensity for racial discrimination (13).

Collaboration between medical professionals, researchers, and AI developers is necessary to establish standardized data collection and reporting methods. Implementing robust data curation processes can help minimize the impact of data discrepancies on the performance of ML algorithms.

5.2. Ethical Considerations and Patient Privacy

While incorporating new technologies is expected in health care, AI-enabled technologies possess characteristics that set them apart from other innovations in ways that can impinge on a therapeutic patient-physician relationship. For example, guidance in the AMA Code of Medical Ethics on ethically sound innovation in medical practice (Opinion 1.2.11) provides that any innovation intended to directly affect patient care be scientifically well grounded and developed in coordination with individuals who have appropriate clinical expertise; that the risks an innovation poses to individual patients should be minimized, and the likelihood that the innovation can be applied to and benefit populations of patients be maximized (12).

The use of AI in medical decision-making raises ethical concerns related to algorithmic bias, transparency, and accountability. For example, attempts to use data from the Framingham Heart Study to predict the risk of cardiovascular events in nonwhite populations have led to biased results, with both overestimations and underestimations of risk (14).

The intent behind the design of machine-learning systems also needs to be considered. Algorithms can be designed to perform in unethical ways. A recent high-profile example is Uber’s software tool Greyball, which was designed to predict which ride hailers might be undercover law-enforcement officers, thereby allowing the company to identify and circumvent local regulations (11). Ensuring patient privacy and data security is crucial in the development and implementation of AI-driven TAVR tools. Regulatory frameworks and guidelines should be established to address these concerns and promote the responsible use of AI in TAVR.

5.3. Integration with Clinical Practice and Education

Successful integration of AI-powered tools into clinical practice requires the development of user-friendly interfaces and seamless integration with existing systems. The tension between improving health and generating profit needs to be acknowledged and addressed especially with the integration of machine learning since the builders and purchasers of machine-learning systems are unlikely to be the same people delivering bedside care (11). Due to this ethical strain, medical professionals must be trained to understand the capabilities and limitations of AI-driven tools, ensuring they can make informed decisions about their use in TAVR. Ongoing research and development efforts are needed to refine and expand the capabilities of AI in TAVR, fostering a collaborative environment between clinicians and AI developers.

6. Conclusion

Machine learning algorithms and artificial intelligence have the potential to significantly augment the safety and predict complications of transcatheter aortic valve replacement. The application of these technologies in patient selection, procedure planning, intraprocedural guidance, and post-procedural monitoring can lead to improved patient outcomes and reduced complications. However, challenges related to data quality and standardization, ethical considerations, and integration with clinical practice must be addressed to fully realize the potential of AI in TAVR. By fostering collaboration between medical professionals, researchers, and AI developers, we can work towards the development of AI-driven tools that will ultimately transform the field of TAVR and improve patient care.

References:

1. Desai ND, O’Brien SM, Cohen DJ, et al. Composite Metric for Benchmarking Site Performance in Transcatheter Aortic Valve Replacement: Results From the STS/ACC TVT Registry. Circulation. 2021;144(3):186-194

2. Evertz R, Lange T, Backhaus SJ, et al. Artificial Intelligence Enabled Fully Automated CMR Function Quantification for Optimized Risk Stratification in Patients Undergoing Transcatheter Aortic Valve Replacement. J Interv Cardiol. 2022;2022:1368878. Published 2022 Apr 20.

3. Willson, AB,. et al. (2012). Computed tomography-based sizing recommendations for transcatheter aortic valve replacement with balloon-expandable valves: Comparison with transesophageal echocardiography and rationale for implementation in a prospective trial. J Cardiovasc Comput Tomogr., 6(6), 406-414.

4. Gohmann, R. F., et al. (2022). Combined cCTA and TAVR Planning for Ruling Out Significant CAD: Added Value of ML-Based CT-FFR. JACC Cardiovasc Imaging, 15(3), 476-486.

5. Kato N, et al. Superiority of novel automated assessment of aortic annulus by intraoperative three-dimensional transesophageal echocardiography in patients with severe aortic stenosis: Comparison with conventional cross-sectional assessment. J Cardiol. 2018;72(4):321-327. doi:10.1016/j.jjcc.2018.02.017

6. Hernandez-Suarez D. F., et al. (2019). Machine Learning Prediction Models for In-Hospital Mortality After Transcatheter Aortic Valve Replacement. JACC Cardiovasc Interv, 12(14), 1328-1338.

7. Hermans MC, Van Mourik MS, Hermens HJ, Baan J Jr, Vis MM. Remote Monitoring of Patients Undergoing Transcatheter Aortic Valve Replacement: A Framework for Postprocedural Telemonitoring. JMIR Cardio. 2018;2(1):e9.

8. Mamprin, M., et al. (2021) Machine Learning for Predicting Mortality in Transcatheter Aortic Valve Implantation: An Inter-Center Cross Validation Study. J Cardiovasc Dev Dis., 8(6), 65.

9. Kwiecinski J et al. Machine learning for prediction of all-cause mortality after transcatheter aortic valve implantation, European Heart Journal – Quality of Care and Clinical Outcomes, 2023

10. Agasthi P, Ashraf H, Pujari SH, et al. Artificial Intelligence Trumps TAVI2-SCORE and CoreValve Score in Predicting 1-Year Mortality Post-Transcatheter Aortic Valve Replacement. Cardiovasc Revasc Med. 2021;24:33-41.

11. Char DS, Shah NH, Magnus D. Implementing Machine Learning in Health Care – Addressing Ethical Challenges. N Engl J Med. 2018;378(11):981-983.

12. American Medical Association. Code of Medical Ethics. Opinion 1.2.1, Ethically sound innovation in medical practice. https://www.ama-assn.org/delivering-care/ethics/ethically-sound-innovation-medical-practice. Accessed 19 Feb 2021.

13. Angwin J, Larson J, Mattu S, Kirchner L. Machine bias. ProPublica. 2016 May 23; ( https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing)

14. Gijsberts CM, Groenewegen KA, Hoefer IE, et al. Race/ethnic differences in the associations of the Framingham risk factors with carotid IMT and cardiovascular events. PLoS One. 2015;10(7):e0132321.

Acknowledgements:

The authors would like to acknowledge the contributions of the medical professionals, researchers, and AI developers working collaboratively to improve patient outcomes and safety in TAVR procedures. The advancements in the field of AI and ML algorithms have the potential to revolutionize patient care, and continued efforts in research and development will help to address the challenges and limitations currently faced in the implementation of these technologies.

Conflicts of Interest:

The authors declare no conflicts of interest.

Funding:

No funding was received for this study.

Author Contributions:

All authors contributed equally to the conception and design of the study, data collection and analysis, and manuscript preparation. All authors have read and approved the final manuscript.

Ethics Approval and Consent to Participate:

Not applicable.

Consent for Publication:

Not applicable.

Availability of Data and Materials:

Not applicable.

Competing Interests:

The authors declare that they have no competing interests.