Researchers at West Virginia University (WVU) are making strides in addressing heart disease diagnosis in rural populations through the development of advanced artificial intelligence (AI) models. This initiative is particularly significant as many existing healthcare AI systems have been criticized for their bias towards urban patient data, potentially sidelining the unique needs of rural communities.
Prashnna Gyawali, an assistant professor in the Benjamin M. Statler College of Engineering and Mineral Resources, explained that AI models are typically trained on data derived from urban settings, often leading to discrepancies in diagnosis for rural patients. “Most of that data comes from affluent urban areas, which differ biologically from rural populations,” Gyawali noted. This gap in data representation has prompted his team to focus on training AI systems using exclusively rural patient data from West Virginia.
Gyawali emphasized the importance of aligning AI algorithms with the populations they aim to serve. “If we want AI models to assist in diagnosing heart disease in our rural population effectively, we must ensure they are trained on the specific characteristics of these populations,” he stated.
To achieve this, the research team has collected anonymous patient datasets from various regions of West Virginia. They are testing different AI models to evaluate their effectiveness in diagnosing heart disease based on medical test results. Gyawali pointed out that if properly implemented, AI can significantly alleviate the burdens faced by rural healthcare systems. It has the potential to not only reduce the workload of healthcare professionals but also facilitate early disease detection, allowing for timely treatment.
“The healthcare challenges are escalating, and we are facing manpower shortages,” Gyawali remarked. “In West Virginia, accessible healthcare infrastructures are limited. Patients may need to travel several hours for initial diagnoses. If we could establish more clinics equipped with affordable scanning devices integrated with AI systems, we could create a robust early detection framework.”
Despite the optimism surrounding the project, Gyawali cautioned that the AI models have so far only interacted with historical data and have not yet been applied to real-world patient scenarios. Continuous refinement of the model is essential until both medical and computer science experts can be assured of its safety and reliability. “In safety-critical applications like healthcare, reliability is paramount,” he explained. “We cannot afford to misdiagnose patients; we must ensure the model accurately identifies those who require immediate attention.”
The research team is committed to enhancing the model’s reliability before it is introduced into clinical trials. While Gyawali did not specify a timeline for these trials, he mentioned that the ongoing process involves adding layers to improve performance. “We are exploring how we can validate these algorithms further,” he stated. “Can we partner with clinics not involved in this study to test our model on their datasets? Our aim is to assess the algorithm’s performance beyond West Virginia and potentially apply it in other states.”
Moreover, Gyawali highlighted the necessity for policy-level interventions to facilitate the introduction of these AI algorithms in real-world clinical settings. “That’s the roadmap to adopting these tools effectively in clinics,” he concluded, underscoring the project’s potential to transform rural healthcare delivery.
As the team continues to develop and refine their AI model, the hope is that it will eventually lead to improved health outcomes for those living in rural areas, making heart disease diagnosis more accurate and accessible.
