Fellowship Support for Ramsey Wehbe

Project: Research project

Project Details


Transthyretin cardiac amyloidosis (ATTR-CA) remains an underdiagnosed cause of heart failure despite remarkable advances in existent diagnostic modalities.1 Early diagnosis of ATTR-CA is critical to ensure favorable patient outcomes given targeted therapy with the ATTR stabilizer tafamidis has demonstrated clinical benefits in improved survival and quality of life.2,3 Non-invasive evaluation with electrocardiography, echocardiography, and magnetic resonance imaging can be helpful in suggesting a diagnosis of ATTR-CA; however, radionuclide imaging with calcium avid radiotracers (cardiac scintigraphy) has emerged as a key component of the diagnostic algorithm for ATTR-CA given its exceptional sensitivity and specificity (up to 95% and 99%, respectively) reported in the literature.4–8 Unfortunately, there are many potential pitfalls in the interpretation of cardiac scintigraphy imaging for ATTR-CA9 and the “real world” discriminative performance may be considerably lower than that reported in the literature, especially at non-referral centers with lower study volumes or prevalence of disease.10 Computer vision systems leveraging artificial intelligence, particularly deep learning systems based on convolutional neural network architectures, have recently been deployed to assist in the interpretation of diagnostic medical imaging with ground-breaking success, often exceeding human-level performance on certain tasks.11–13 We hypothesize that a deep learning system trained to classify cardiac scintigraphy studies as positive or negative for ATTR-CA would improve discriminative performance and increase interrater reliability in the diagnosis of ATTR-CA, even at less experienced centers. Such a system could enable earlier and more accurate detection of ATTR-CA to ensure that patients who might benefit from therapy with tafamidis receive a timely diagnosis. The objectives of this study are to a) assess the discriminative performance of cardiac scintigraphy across different center volumes, b) train a deep learning computer vision system to assist in the diagnosis of ATTR-CA on cardiac scintigraphy imaging, and c) prospectively evaluate the discriminative performance of this model and compare to the performance of board-certified nuclear cardiologists with and without the assistance of the deep learning model. In the long term, we feel this research has the potential to significantly improve clinical outcomes for patients with ATTR-CA by precluding more invasive evaluation (e.g. endomyocardial biopsy) and more accurately identifying patients that might benefit from early initiation of therapy.
Effective start/end date10/1/219/30/22


  • Pfizer Inc. (NOT SPECIFIED)


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.