In a significant advancement for cardiac healthcare, The Lancet Digital Health has recently published a commentary article detailing an artificial intelligence (AI)-driven approach to improve the screening for transthyretin (ATTR) cardiac amyloidosis.
This condition, often diagnosed in older patients with heart failure or aortic stenosis, has a notably poor prognosis if left untreated. However, early intervention with targeted therapy can substantially reduce mortality and hospitalizations related to cardiovascular issues.
The study, conducted by Clemens P. Spielvogel and colleagues, presents a novel multistep method for large-scale, opportunistic screening using bone scintigraphy imaging. The process begins with normalizing image counts, followed by employing deep learning techniques to crop images of the thorax. These cropped images are then used as inputs for a convolutional neural network (CNN), which classifies the images based on the presence of abnormal planar uptake.
The model’s predictive performance was outstanding, boasting an area under the curve (AUC) of 1.000 during internal cross-validation and maintaining consistent results across different cohorts and radiotracers with AUCs ranging from 0.925 to 1.0001. The approach was further refined to address common sources of false positives and negatives, enhancing the positive predictive value from 0.886 to 0.932.
This AI model stands out not only for its large patient population but also for its rigorous external testing across multiple centers with diverse imaging protocols. Such external validation is crucial as it assesses the model’s performance in new populations and under varying conditions.
The ground truth for the study was established through a core laboratory approach involving triple reviews of each study and additional reviews for discrepancies. This meticulous validation process underscores the reliability of the findings and the potential of AI to revolutionize the early detection of ATTR-cardiac amyloidosis.
The research by Spielvogel and team marks a pivotal step in cardiac health, offering a generalizable and efficient solution for screening a disease that has historically been challenging to diagnose promptly. As AI continues to integrate into healthcare, it promises to bridge critical care gaps and enhance patient outcomes.
