West Virginia University researchers are creating artificial intelligence models to improve the diagnosis and prediction of heart disease among rural patients. This initiative is particularly important as many existing AI frameworks in healthcare predominantly reflect urban populations, which may not accurately represent rural demographics.
According to Prashnna Gyawali, an assistant professor at the university”s Benjamin M. Statler College of Engineering and Mineral Resources, current AI models are typically trained on data derived from urban settings. These settings are often more affluent, and the biological characteristics of patients in these areas can differ significantly from those in rural communities. This discrepancy hampers the effectiveness of AI systems in rural healthcare environments.
To address this issue, Gyawali and his team have initiated the development of 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 emphasized that for AI to efficiently assist in diagnosing heart disease and other medical conditions among rural populations, the models must reflect the specific characteristics and needs of these communities.
The team has gathered anonymous patient data from various regions across West Virginia and is employing this information to evaluate different AI models for their diagnostic capabilities concerning heart disease. Gyawali highlighted that when properly implemented, AI could greatly enhance rural healthcare systems. It can alleviate some of the burdens faced by overworked medical staff and facilitate the early detection of health issues, allowing for timely interventions.
“Healthcare challenges are escalating while we are experiencing a shortage of healthcare personnel,” Gyawali noted. He pointed out that in West Virginia, accessible healthcare facilities are limited, and patients may need to travel several hours for initial assessments. He envisions a future where more clinics equipped with affordable scanning technology integrated with AI could provide early detection systems for at-risk patients.
Gyawali stressed the importance of ensuring that the AI models are trained on local populations, asserting that reliability and impartiality are paramount. Although the testing of the models has been promising so far, he cautioned that they have only interacted with historical rural datasets and have not yet been applied to real-world patient scenarios. Ongoing refinement of the AI model is essential until both medical and computer science experts are assured of its safety and reliability for human use.
“In safety-critical areas like healthcare, we must ensure the technology is dependable,” he remarked. “We do not want to misdiagnose patients or overlook those in need of urgent care. If AI serves as a component of the healthcare system, it must perform its role effectively.”
The research team is committed to further developing the AI model to enhance its reliability prior to its use in clinical trials. While there is no set timeline for these trials, Gyawali indicated that the process is progressing. “We are adding more layers to ensure the model is dependable. How can we improve its performance? These are the questions we are exploring in our lab. We are also considering testing the algorithm in clinics outside of this study to evaluate its performance in different settings,” he explained.
He also expressed the need for policy-level interventions to facilitate the implementation of these algorithms in real-world clinical environments. “This is the roadmap for integrating these tools into clinics,” Gyawali concluded.
