Quantitative Prediction of Disease and Outcomes from Next Generation SPECT and CT Coronary artery disease remains a major public health problem worldwide. It causes approximately 1 of every 6 deaths in the United States. Imaging of myocardial perfusion (delivery of blood to the heart muscle) by myocardial perfusion single photon emission tomography (MPS) allows physicians to detect disease before heart attacks occur and is currently used to predict risk in millions of patients annually. Under the current grant, we have established a unique collaborative multicenter registry including over 23,000 imaging datasets (REFINE SPECT) with both prognostic (major adverse cardiovascular events) and diagnostic (invasive catheterization) outcomes. Using this registry, we have demonstrated that a combination of MPS image analysis and artificial intelligence (AI) tools achieved superior predictive performance compared to visual assessment by experienced readers or current state-of-the-art quantitative techniques. In the renewal, we plan to expand REFINE SPECT with now-available enhanced datasets (adding CT and myocardial blood flow information) and leverage latest AI advances to provide a personalized decision support tool for patient-specific cardiovascular risk assessment and estimation of benefit from revascularization following MPS. The overall aim is to optimize the clinical capabilities of MPS in risk prediction and treatment guidance by integrating all available imaging and clinical data with state-of-the-art AI methods. For this work, we propose the following 3 specific aims: (1) To expand and enhance our REFINE SPECT registry including CT and MPS flow data, (2) To develop fully automated techniques for all MPS and CT image analysis, (3) To apply explainable deep learning time-to-event AI models for optimal prediction of MACE and benefit from revascularization from all image and clinical data. This work will result in an immediately deployable clinical tool, which will optimally predict risk of adverse events and establish the relative benefits from specific therapies, beyond what is possible by subjective visual analysis and mental integration of all imaging (MPS, CT, flow), and clinical data by physicians. Such quantitative integrative methods are not yet available, leaving the current practice for assessing risk and recommending therapy highly subjective. The precise quantitative results will be presented to clinicians in easy to understand terms (e.g., % risk per year, or relative risk of one therapy vs. the alternative) for a specific patient. Additionally, our methods to make AI conclusions more tangible will improve adoption of this technology. All results will be derived fully automatically thus eliminating any variability. Our approach will fit into current MPS practice and will be immediately translatable to clinics worldwide. Most importantly, this research will allow patients to benefit from increased precision and accuracy in risk assessment, thereby optimizing the use of imaging in guiding patient management decisions and ultimately improving outcomes.
Myocardial perfusion imaging with SPECT is often used to predict who is at risk of heart attack and should undergo treatment such as coronary bypass or stenting; however, physicians read images visually and report results with wide variability. With the latest artificial intelligence tools and new types of imaging (including CT and fast SPECT scans), the investigators propose to develop and validate an automated clinical tool to optimize risk prediction and objectively establish the relative benefit of a specific therapy. This new tool will consider all available patient images and other relevant information to provide a personalized explanation and precise calculation of risk and potential benefits from therapy for each patient.
|Sharif, Behzad; Motwani, Manish; Arsanjani, Reza et al. (2018) Impact of incomplete ventricular coverage on diagnostic performance of myocardial perfusion imaging. Int J Cardiovasc Imaging 34:661-669|
|Betancur, Julian; Otaki, Yuka; Motwani, Manish et al. (2018) Prognostic Value of Combined Clinical and Myocardial Perfusion Imaging Data Using Machine Learning. JACC Cardiovasc Imaging 11:1000-1009|
|Motwani, Manish; Leslie, William D; Goertzen, Andrew L et al. (2018) Fully automated analysis of attenuation-corrected SPECT for the long-term prediction of acute myocardial infarction. J Nucl Cardiol 25:1353-1360|
|Gomez, Javier; Doukky, Rami; Germano, Guido et al. (2018) New Trends in Quantitative Nuclear Cardiology Methods. Curr Cardiovasc Imaging Rep 11:|
|Betancur, Julian; Commandeur, Frederic; Motlagh, Mahsaw et al. (2018) Deep Learning for Prediction of Obstructive Disease From Fast Myocardial Perfusion SPECT: A Multicenter Study. JACC Cardiovasc Imaging 11:1654-1663|
|Išgum, Ivana; de Vos, Bob D; Wolterink, Jelmer M et al. (2018) Automatic determination of cardiovascular risk by CT attenuation correction maps in Rb-82 PET/CT. J Nucl Cardiol 25:2133-2142|
|Betancur, Julian A; Hu, Lien-Hsin; Commandeur, Frederic et al. (2018) Deep Learning Analysis of Upright-Supine High-Efficiency SPECT Myocardial Perfusion Imaging for Prediction of Obstructive Coronary Artery Disease: A Multicenter Study. J Nucl Med :|
|Haro Alonso, David; Wernick, Miles N; Yang, Yongyi et al. (2018) Prediction of cardiac death after adenosine myocardial perfusion SPECT based on machine learning. J Nucl Cardiol :|
|Slomka, Piotr J; Betancur, Julian; Liang, Joanna X et al. (2018) Rationale and design of the REgistry of Fast Myocardial Perfusion Imaging with NExt generation SPECT (REFINE SPECT). J Nucl Cardiol :|
|Chaudhry, Waseem; Hussain, Nasir; Ahlberg, Alan W et al. (2017) Multicenter evaluation of stress-first myocardial perfusion image triage by nuclear technologists and automated quantification. J Nucl Cardiol 24:809-820|
Showing the most recent 10 out of 67 publications