An innovative artificial intelligence system crafted by researchers at The Ohio State University has shown remarkable efficiency in forecasting the most suitable stroke prevention therapies for patients with cardiac ailments, matching or surpassing the prediction capabilities of traditional randomized clinical studies.
The findings, unveiled in the “Patterns” journal on May 1, 2024, suggest that this AI might transform how medical professionals tailor treatment options for their patients.
Using a “foundation model” approach, the AI leveraged a vast array of anonymized patient records from numerous health institutions. “Current algorithms fall short of performing this task,” stated Ping Zhang, the study’s principal investigator and associate professor at Ohio State with expertise in computer science, engineering, and biomedical informatics. Zhang observed that their method’s precision improved by 7% to 8% compared with existing techniques.
Forging Ahead with Tailored Patient Treatment
The researchers are optimistic that this technology could expedite clinical trials and decrease their costs, thereby enhancing the customization of care for patients. Ruoqi Liu, a doctoral student who took part in the research, commented, “Our model might function as a quickening component, pinpointing a narrow selection of likely effective drugs for beating a disease.”
CURE is the acronym given to their method, meaning CaUsal tReatment Effect estimation. The team highlighted its adaptability for a wide array of diseases and medications. After initial modifications with specific details of diseases and treatments, Liu indicated that their model is capable of immediate recalibration for varying tasks.
Their approach to crafting the AI involved the combination of patient records and biomedical knowledge into an advanced foundation model that harmonizes data with contextual understanding. Liu said, “Through KG-TREAT, our knowledge-enhanced foundation model, we aim to blend patient information with knowledge graphs optimally, ensuring the model comprehends patient data more effectively.”
The predictive accuracy of this AI against actual clinical trial results was examined by the researchers, confirming its high effectiveness due to its comprehensive pre-training and further boosted by the integration of knowledge graphs. Zhang contemplates the eventual sanctioning of AI like theirs by the FDA as an assistive tool for decision-making purposes. This AI would empower clinicians in crafting individualized treatment plans.
The project, supported by the National Institutes of Health, included collaborative efforts from Ping-Yu Chen of IBM Research and Lingfei Wu of Anytime AI. This pioneering investigation holds the promise of a fundamental shift in the mode of strategizing treatments in the medical field, paving the way to a future where patient care is exceedingly customized..
