Researchers have developed an artificial intelligence-based tool to diagnose transthyretin amyloid cardiomyopathy (ATTR-CM) using total body scintigraphy imaging, according to a study recently published in the European Journal of Nuclear Medicine and Molecular Imaging.
“AI-based image analysis provides automated tools for more accurate detection and quantification of ATTR-CM from imaging studies to address the existing issues in scintigraphy and [single photon emission computed tomography] image analysis,” the authors said.
What is ATTR-CM?
Transthyretin amyloidosis cardiomyopathy (ATTR-CM) is a rare progressive disease of the heart muscle that leads to congestive heart failure. It occurs when the transthyretin protein produced by the liver is unstable. Symptoms include fatigue; shortness of breath; irregular heart rate or palpitations; swelling of the legs, ankles and stomach; brain fog; wheezing; and dizziness. It often goes underdiagnosed because of a lack of awareness and knowledge of the disease. There is currently no cure for ATTR-CM.
Because ATTR-CM shares symptoms with several other cardiovascular diseases and has no standardized diagnostic procedure, the disease may sometimes go undiagnosed. Scintigraphy is emerging as the preferred method of diagnosing patients with ATTR-CM, but it is susceptible to human error.
Read more about ATTR-CM testing and diagnosis
In this study, the authors used six datasets from multiple centers to develop a deep learning system for detecting cases of ATTR-CM and classifying the cases based on their severity.
The tool was trained first to locate the heart, then determine the level of cardiac amyloidosis present in the individual. One of the datasets did not have prior information of the presence of cardiac amyloidosis, allowing the authors to evaluate the screening capability of the model.
The investigators found that their models for detecting and scoring ATTR-CM performed well, both in the datasets that were used to develop the model and in multiple external datasets. The models were able not only to accurately and consistently identify the presence of ATTR-CM in these images, but also to reliably categorize the disease severity.
Furthermore, the study analyzed 3,215 images to assess the utility of the tool in disease screening. In total, 10 cases were identified as potentially positive for ATTR-CM. When received by a nuclear medicine physician, four of the cases were confirmed.
“This fully automated pipeline could lead to timely and accurate diagnosis, ultimately improving patient outcomes,” the study concluded.
Sign up here to get the latest news, perspectives and information about ATTR-CM sent directly to your inbox. Registration is free and takes only a minute.
