The NRC is teaming up with Montreal's Jewish General Hospital to study machine learning tools for health care.
Technology is an integral part of modern medicine. However, the health care system is still under a heavy burden, and places in intensive care have been all too precious since the beginning of the pandemic. What if we could take some load off the system using the latest technologies to come out of the computer science world? Technology that could advise doctors on what decisions to make. Machine learning (ML) has already revolutionized countless industries, and it is poised to do so in the medical community as well. For this to happen, though, doctors and other health experts will need to know that they can trust these new computer tools to help them make decisions that will have a real impact in their patients' well-being.
The National Research Council of Canada (NRC), in collaboration with the CIUSSS West-Central Montreal's digital health team and OROT—the Connected Health Innovation Hub at the Jewish General Hospital, is currently working on exactly such a project.
Its goal: better integration of ML algorithms into a hospital's decision-making process by providing explanations along with the algorithm's output.
This project is focussed on explainable ML, which is a set of techniques that can learn from a wide range of data in order to draw conclusions from subsequent data. For example, by analyzing the information entered by health care staff when triaging COVID-19 patients in an emergency ward, these methods could be used to help predict which patient is most likely to require specialized care in the coming days. Explainable ML can go one step further by helping to clarify the reason for the predictions it makes in a way that the doctor can understand. When combined with a doctor's experience, such a tool could assist the hospital staff to make quick, sensible and accurate decisions.
ML offers interesting possibilities for the future of our health care system. These ML algorithms can be faster and more accurate than humans, and incorporate enormous amounts of data. The main objective of this research project is to understand how ML can best be adopted by physicians to become an essential support tool in their health care practice.
As ML is a relatively new technology in the health sector, the NRC's Medical Devices Research Centre's Jordan Hovdebo, Team Lead responsible for the project, and Rola Harmouche, Project Lead, have planned the project in several research phases.
The team will first work with clinicians to identify the issues that need to be addressed during a pandemic—although once finalized, these methods could be used in a number of other health care situations. Next, a screening of existing research and technologies will be carried out to find the ones that will be most useful in these situations. The research team will then develop and adapt an algorithm to make the right predictions in an explainable way. These algorithms will then be combined with user experience and interface guidelines for clinical informatics systems, to be developed by researchers at the Jewish General Hospital, into a prototype that demonstrates the explainable approach to ML. The prototype will then be presented for evaluation by the participating team of clinicians.
"We want to build something that will have a long-term impact in clinical settings."
Started in February 2021, the first phase of the project is planned to last just over a year. At this stage, development and validation of the ML methods will be based on publicly available data. However, the tool may be adapted to use data from cases at the Jewish General Hospital in the latter parts of the collaboration. If all goes as planned, this project will produce a validated prototype that demonstrates the efficacy of the approach which could be used in actual products to improve health care systems everywhere.
This project is being funded by the NRC Pandemic Response Challenge program. This program's goal is to develop technology and solutions that will fill the gaps in our medical community's ability to respond to a pandemic. Although a crisis like COVID-19 can severely disrupt our activities, it also brings about innovative spirit and will produce technologies that will change health care for the better.
Science or science fiction?
Imagine a doctor having a discussion with his computer about a patient’s health and it is providing recommendations for care. It may seem far-fetched, but the same kinds of technology that are powering your internet search function and your bank’s fraud detection systems are starting to find their place in your local hospital. The NRC is working with Montreal’s Jewish General Hospital to test ways of interacting with and presenting technologies like these to make sure they meet the needs of the health care community.