AI Fast-Tracks Disease Risk Predictions

Recent advancements in artificial intelligence offer a groundbreaking opportunity to proactively assess the likelihood of developing severe health conditions later in life with a simple touch of a button.

Abdominal aortic calcification (AAC) refers to the accumulation of calcified deposits within the abdominal aorta’s walls, serving as a crucial predictor of cardiovascular disease events, including heart attacks and strokes. Additionally, AAC assessment can effectively gauge the risks of falls, fractures, and late-life dementia.

The remarkable aspect is that the prevalent bone density scanning machines utilized for osteoporosis detection can also identify AAC. Nevertheless, the analysis of these images necessitates highly trained experts, resulting in a time-consuming process of 5 to 15 minutes per image.

Addressing this challenge, a collaborative effort between Edith Cowan University’s (ECU) School of Science and School of Medical and Health Sciences has yielded an innovative software solution capable of significantly expediting the analysis of scans. Astonishingly, this cutting-edge software can assess approximately 60,000 images in a single day.

Associate Professor Joshua Lewis, a prominent researcher and Heart Foundation Future Leader Fellow, emphasizes the pivotal role of this efficiency boost in enabling widespread AAC utilization for research and aiding individuals in preventing future health complications. He envisions the potential for early cardiovascular disease detection and continuous disease monitoring during routine clinical practice by swiftly acquiring images and automated scores during bone density testing.

The research findings are the result of an extensive international collaboration between ECU, the University of WA, University of Minnesota, Southampton, University of Manitoba, Marcus Institute for Aging Research, and Hebrew SeniorLife Harvard Medical School—a testament to the multidisciplinary and global nature of the study.

While not the initial algorithm developed for AAC assessment based on these images, this study represents the most extensive endeavor of its kind, employing the most commonly used bone density machine models, and implementing real-world testing using images obtained from routine bone density evaluations. Expert analysis and the team’s software meticulously examined over 5,000 images, ultimately reaching consistent conclusions regarding the extent of AAC (low, moderate, or high) 80% of the time. These results are particularly remarkable given that it was the software’s first version.

Crucially, the software misdiagnosed only 3% of individuals with high AAC levels as having low levels. Professor Lewis underscores the significance of this achievement since this group exhibits the most severe disease progression and faces the highest risks of fatal and nonfatal cardiovascular events, as well as overall mortality.

Professor Lewis further emphasizes that while enhancements are still required to refine the software’s accuracy compared to human readings, substantial improvements have already been achieved with subsequent software versions. The ability to automatically assess the presence and severity of AAC with comparable accuracy to imaging specialists paves the way for large-scale screening of cardiovascular disease and other conditions, even in individuals without apparent symptoms.

Ultimately, this revolutionary development empowers individuals at risk to adopt necessary lifestyle changes at an earlier stage, positioning them for improved health and well-being in their later years.

The Heart Foundation’s support, facilitated by Professor Lewis’ 2019 Future Leadership Fellowship, has been instrumental in funding this project, enabling dedicated research over a three-year period.

The comprehensive research findings are documented in the publication “Machine Learning for Abdominal Aortic Calcification Assessment from Bone Density Machine-Derived Lateral Spine Images,” available in eBioMedicine.


Source ScienceDaily

Author: Neurologica

Leave a Reply