Utilizing the Power of Artificial Intelligence for Stroke Prevention
A team from The Ohio State University has made significant strides in stroke prevention measures for heart disease patients with the use of artificial intelligence. Their innovative AI strategy harbors the potential to redefine clinical trial execution.
A Groundbreaking AI Algorithm
The team has developed an AI algorithm that excels at predicting the most beneficial interventions to prevent strokes. Its effectiveness could potentially surpass the accuracy associated with randomized clinical trials, a gold standard in medical research. The algorithm uses anonymized patient data obtained from healthcare claims to provide personalized treatment plans, eliminating the need for physical clinical trials.
Introducing the ‘Digital Twin’
The researchers from Ohio State propose a major shift in clinical decision-making with their concept of a ‘digital twin’ for each patient. This virtual model grants healthcare professionals the ability to identify the optimal treatment paths with higher precision. The model operates similarly to generative AI applications, such as ChatGPT, using comprehensive patient records to estimate stroke risk and predict potential treatments.
Unique and Efficient
Leading the study, Ping Zhang, an associate professor in both computer science, engineering, and biomedical informatics, highlights the unique efficiency of their algorithm. Zhang emphasizes their AI’s ability to replicate the results of a randomized clinical trial, outperforming competing models by a notable 7% to 8%.
The CURE Framework
The CURE framework, which stands for CaUsal tReatment Effect estimation, marks a breakthrough in using healthcare claim records in conjunction with biomedical knowledge graphs. This combination allows the AI to quickly adjust and provide accurate prognoses across a multitude of medical conditions and treatments.
In the pre-training phase, the researchers utilized data from the MarketScan Commercial Claims and Encounters, analyzing a sample of 3 million patient cases encompassing a variety of medical and medication codes.
Ruoqi Liu, a doctoral candidate for computer science and engineering in Zhang’s lab, played a crucial role in the model’s development. Liu applauded the model’s adaptability, demonstrating an ability to pre-train on large-scale datasets without needing specific treatments.
The implications of this research extend to speeding up drug selection in clinical trials, increasing the accuracy of personalized therapy, and reducing the costs and time traditionally associated with clinical research. The project showcases the growing influence of AI in healthcare decision-making. Awaiting approval from regulatory agencies such as the Food and Drug Administration, this study serves as a beacon of hope regarding the future integration of AI into healthcare.
The research was published in the journal Patterns on May 1, 2024, and was supported by the National Institutes of Health. The study was also underpinned by contributions from esteemed individuals such as Pin-Yu Chen from IBM Research and Lingfei Wu of Anytime AI.
