A moment of profound loss set Blessing Ogbuokiri on a research journey that could change, and perhaps even save, people’s lives.

After his mother passed due to a heart issue, the Brock University Assistant Professor of Computer Science was inspired to research the use of artificial intelligence to support others facing similar health challenges.

Ogbuokiri is developing and evaluating a machine learning model that could potentially assess whether older heart patients are likely to be admitted to hospital or die because of heart failure.

“I’m not a medical doctor, but I feel I can contribute to proffering a solution that potentially prevents heart failure using my knowledge of artificial intelligence,” says the Director of Brock’s Responsible and Applied Machine Learning Laboratory (RAML Lab).

He and his student research team have received funding from Brock University’s new Black Scholar Research Grant for their work.

“It is particularly fitting that this grant is supporting Dr. Ogbuokiri in pursuing a project informed by his personal history and so incredibly valuable to Niagara’s aging population,” says Acting Vice-President, Research Michelle McGinn. “His enthusiasm for helping patients combined with his technical expertise and his commitment to honouring the memory of his mother embody the spirit of the Black Scholar Research Grant.”

Ogbuokiri and his team are training a machine learning model to make associations between a range of variables based on health data from the Canadian Longitudinal Study on Aging. These include medical history, smoking status, physical activity level, socioeconomic status and the presence of chronic conditions like diabetes.

“When we have trained it and it’s smart enough to recognize all these patterns, the model can give us a prediction and say, for example, there’s a likelihood this person has a 50 per cent chance of suffering heart failure and being admitted to hospital,” says Ogbuokiri.

He says the team aims to create a tool that patients and health-care professionals can use to assess heart failure risk “at the click of a button.”

This knowledge could motivate patients to make healthy lifestyle changes, such as exercising or quitting smoking, he says, or help health-care professionals take proactive measures to prevent the health system from being overwhelmed with large numbers of admissions at once.

Ogbuokiri says his model could also improve access to early interventions for patients from Black and equity-seeking communities, who are disproportionately affected by heart failure and can face biases and other barriers to receiving health care.

These same stereotypes and harmful prejudices can be embedded into machine learning models during the training process, resulting in unequal opportunities, distorted information and other negative impacts.

In health-care models, this may take the form of systematically underpredicting risk for certain populations, such as Black or low-income patients, leading to disparities in access to timely interventions or treatments.

Ogbuokiri says the researchers are working to avoid and mitigate biases in their model by “applying bias mitigation techniques during data preprocessing and evaluating model fairness using metrics to ensure equitable performance across demographic groups.”