Improved accuracy of myocardial perfusion SPECT for detection of coronary artery disease by machine learning in a large population

• Published in: Journal of nuclear cardiology Vol. 20; no. 4; pp. 553 - 562
• Main Authors: Arsanjani, Reza, Xu, Yuan, Dey, Damini, Vahistha, Vishal, Shalev, Aryeh, Nakanishi, Rine, Hayes, Sean, Fish, Mathews, Berman, Daniel, Germano, Guido, Slomka, Piotr J.
• Format: Journal Article
• Language: English
• Published: Boston Elsevier Inc 01.08.2013; Springer US; Springer Nature B.V
• Subjects: Aged; Algorithms; Artificial Intelligence; automated quantification; Cardiology; Coronary Angiography - methods; coronary artery disease; Coronary Artery Disease - diagnostic imaging; Electrocardiography - methods; Female; Humans; Imaging; machine learning; Male; Medicine; Medicine & Public Health; Middle Aged; Myocardial Ischemia - pathology; Myocardial Perfusion Imaging - methods; Myocardial perfusion imaging: SPECT; Nuclear Medicine; Observer Variation; Original Article; Pattern Recognition, Automated; Perfusion; Radiology; Radiopharmaceuticals; Reproducibility of Results; ROC Curve; Technetium Tc 99m Sestamibi; Tomography, Emission-Computed, Single-Photon - methods; total perfusion deficit; coronary artery disease; automated quantification; total perfusion deficit; Myocardial perfusion imaging: SPECT; machine learning
• ISSN: 1071-3581; 1532-6551; 1532-6551
• DOI: 10.1007/s12350-013-9706-2

We aimed to improve the diagnostic accuracy of myocardial perfusion SPECT (MPS) by integrating clinical data and quantitative image features with machine learning (ML) algorithms. 1,181 rest 201Tl/stress 99mTc-sestamibi dual-isotope MPS studies [713 consecutive cases with correlating invasive coronary angiography (ICA) and suspected coronary artery disease (CAD) and 468 with low likelihood (LLk) of CAD <5%] were considered. Cases with stenosis <70% by ICA and LLk of CAD were considered normal. Total stress perfusion deficit (TPD) for supine/prone data, stress/rest perfusion change, and transient ischemic dilatation were derived by automated perfusion quantification software and were combined with age, sex, and post-electrocardiogram CAD probability by a boosted ensemble ML algorithm (LogitBoost). The diagnostic accuracy of the model for prediction of obstructive CAD ≥70% was compared to standard prone/supine quantification and to visual analysis by two experienced readers utilizing all imaging, quantitative, and clinical data. Tenfold stratified cross-validation was performed. The diagnostic accuracy of ML (87.3% ± 2.1%) was similar to Expert 1 (86.0% ± 2.1%), but superior to combined supine/prone TPD (82.8% ± 2.2%) and Expert 2 (82.1% ± 2.2%) (P < .01). The receiver operator characteristic areas under curve for ML algorithm (0.94 ± 0.01) were higher than those for TPD and both visual readers (P < .001). The sensitivity of ML algorithm (78.9% ± 4.2%) was similar to TPD (75.6% ± 4.4%) and Expert 1 (76.3% ± 4.3%), but higher than that of Expert 2 (71.1% ± 4.6%), (P < .01). The specificity of ML algorithm (92.1% ± 2.2%) was similar to Expert 1 (91.4% ± 2.2%) and Expert 2 (88.3% ± 2.5%), but higher than TPD (86.8% ± 2.6%), (P < .01). ML significantly improves diagnostic performance of MPS by computational integration of quantitative perfusion and clinical data to the level rivaling expert analysis.