Machine Learning for Predicting Mortality in Transcatheter Aortic Valve Implantation: An Inter-Center Cross Validation Study.

Marco Mamprin,Peter H N De With,Svitlana Zinger,Martijn S Van Mourik,Jo M Zelis,Pim A L Tonino,Bas A J M De Mol,Marije M Vis,Ricardo R Lopes

doi:10.3390/jcdd8060065

Abstract

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. Despite the advancement of machine learning in healthcare, data sharing regulations are very strict and typically prevent exchanging patient data, without the involvement of ethical committees. A very robust validation approach, including 1300 and 631 patients per center, was performed to validate a machine learning model of one center at the other external center with their data, in a mutual fashion. This was achieved without any data exchange but solely by exchanging the models and the data processing pipelines. A dedicated exchange protocol was designed to evaluate and quantify the model’s robustness on the data of the external center. Models developed with the larger dataset offered similar or higher prediction accuracy on the external validation. Logistic regression, random forest and CatBoost lead to areas under 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. We propose a scalable exchange protocol which can be further extended on other TAVI centers, but more generally to any other clinical scenario, that could benefit from this validation approach.

Highlights

We performed two external validations for two transcatheter aortic valve implantation (TAVI) centers, which have been achieved by exchanging the models rather than exchanging the patient data
Assessing missing values and feature distributions is essential for both analyzing the main differences across the two populations, and for a careful selection of the clinical features that can be used in the joint modeling
We have proven that the prediction accuracy of models generated at one TAVI center can be extended to another TAVI center

Summary

Introduction

Aortic valve stenosis (AS) is one of the most common valve diseases in the developed world, primarily impacting the elderly population [1]. The disease has a devastating impact when left untreated, rapidly increasing the risk for heart failure, which can lead to death in many cases. The most common cause of AS is calcification, resulting from a network of pathological processes such as endothelial dysfunction and injury, inflammation, synthetic and osteogenic dedifferentiation, pro-fibrotic response of valvular interstitial cells and degradation of the extracellular matrix by proteolytic enzymes. The calcium deposits on the valve leaflets result in impaired leaflet motion and subsequent obstruction and/or regurgitation of blood flow.

Objectives

Methods

Results

Discussion

Conclusion