West Virginia University (WVU) researchers are creating artificial intelligence models aimed at improving the diagnosis and prediction of heart disease among rural patients. This initiative responds to a significant issue: many existing AI models in healthcare are primarily based on data from urban populations, leading to potential biases that can hinder effective treatment in rural areas.
Prashnna Gyawali, an assistant professor in the WVU Benjamin M. Statler College of Engineering and Mineral Resources, emphasized the importance of using data that accurately reflects rural populations. He noted that most AI diagnostic tools are trained on datasets from affluent urban environments, which differ biologically from rural patients. This discrepancy can compromise the effectiveness of AI systems in rural healthcare settings.
To address this gap, Gyawali and his team have focused on developing a new AI model that exclusively utilizes data from rural patients in West Virginia. “You have to ensure your algorithms have seen the populations where you want them applied,” Gyawali stated. He believes that for AI to effectively assist in diagnosing heart disease in rural communities, it is essential that these models are trained on relevant demographic characteristics.
The research team has gathered anonymous patient datasets from various regions within West Virginia. They are currently testing different AI models to assess their ability to diagnose heart disease based on specific medical test results. Gyawali pointed out that if successful, AI could significantly benefit rural healthcare systems by alleviating the workload of healthcare professionals and enabling early detection of conditions like heart disease.
“Healthcare problems are growing and we have manpower shortages,” Gyawali explained. In West Virginia, access to healthcare can be limited, with patients often having to travel several hours for a proper diagnosis. By implementing AI systems in local clinics equipped with affordable scanning devices, the potential for early detection of health issues could drastically improve.
Despite the promising direction of this research, Gyawali cautioned that the AI models are currently only interacting with historical datasets and have yet to be tested on real-world patients. Ongoing refinement of the model is critical to ensure that it meets safety and reliability standards before being utilized in clinical settings.
“Whenever we talk about safety-critical applications like healthcare, we need to make sure they’re reliable,” he said. Gyawali stressed the importance of accurately identifying patients who require immediate care to avoid misdiagnoses. If integrated effectively, AI could serve as a valuable tool within the healthcare system.
The team is committed to enhancing the AI model, exploring various methods to improve its performance. “How can we further enhance performance? These are AI questions my lab is trying to answer,” Gyawali stated. He also mentioned the potential for testing the AI model in clinics outside West Virginia to assess its generalizability.
Looking forward, Gyawali underscored the need for policy-level interventions to facilitate the introduction of these algorithms in real-world clinical settings. “That’s the roadmap toward adopting these tools in clinics,” he concluded. As the research progresses, the team aims to ensure that the technology can reliably serve rural communities and advance healthcare accessibility.
