Algorithm for predicting valvular heart disease from heart sounds in an unselected cohort - From research project to patent and beyond
Presenters: Lars Ailo Bongo(1,2), Per Niklas Waaler(1) and Ingrid Skjæveland(3)
1 Department of Computer Science, UiT
2 SFI Visual Intelligence
We have developed and patented a machine learning algorithm to detect valvular heart disease using heart sounds recorded in Tromsøundersøkelsen. Despite the cohort being unselected, the algorithm detected aortic stenosis with performance exceeding performance achieved in similar studies based on selected cohorts. Detection of aortic and mitral regurgitation based on heart sound audio was unreliable, but sensitivity was considerably higher for symptomatic cases, and inclusion of clinical variables improved prediction significantly.
In this seminar, we will first present the algorithm and an evaluation of its performance. Then we will describe our lessons learned patenting our research results, and our future plans for commercializing the results. Finally, we will describe how we combine protection through patents with open science principles.
Pre-print of our paper: https://www.medrxiv.org/content/10.1101/2022.11.28.22279153v1
Open source code: https://github.com/uit-hdl/heart-sound-classification
Description of variables http://tromsoundersokelsen.uit.no/tromso/