Static CT myocardial perfusion imaging: image quality, artifacts including distribution and diagnostic performance compared to 82Rb PET

Background Rubidium-82 positron emission tomography (82Rb PET) MPI is considered a noninvasive reference standard for the assessment of myocardial perfusion in coronary artery disease (CAD) patients. Our main goal was to compare the diagnostic performance of static rest/ vasodilator stress CT myocardial perfusion imaging (CT-MPI) to stress/ rest 82Rb PET-MPI for the identification of myocardial ischemia. Methods Forty-four patients with suspected or diagnosed CAD underwent both static CT-MPI and 82Rb PET-MPI at rest and during pharmacological stress. The extent and severity of perfusion defects on PET-MPI were assessed to obtain summed stress score, summed rest score, and summed difference score. The extent and severity of perfusion defects on CT-MPI was visually assessed using the same grading scale. CT-MPI was compared with PET-MPI as the gold standard on a per-territory and a per-patient basis. Results On a per-patient basis, there was moderate agreement between CT-MPI and PET-MPI with a weighted 0.49 for detection of stress induced perfusion abnormalities. Using PET-MPI as a reference, static CT-MPI had 89% sensitivity (SS), 58% specificity (SP), 71% accuracy (AC), 88% negative predictive value (NPV), and 59% positive predictive value (PPV) to diagnose stress-rest perfusion deficits on a per-patient basis. On a per-territory analysis, CT-MPI had 73% SS, 65% SP, 67% AC, 90.8% NPV, and 34% PPV to diagnose perfusion deficits. Conclusions CT-MPI has high sensitivity and good overall accuracy for the diagnosis of functionally significant CAD using 82Rb PET-MPI as the reference standard. CT-MPI may play an important role in assessing the functional significance of CAD especially in combination with CCTA.

Deep Learning Analysis of Polar Maps from SPECT Myocardial Perfusion Imaging for Prediction of Coronary Artery Disease

Purpose: This study aimed to investigate the diagnostic accuracy of deep convolutional neural networks for classifying the polar map images in Single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) by considering the physician’s diagnosis as reference.Methods: 3318 images of stress and rest polar maps related to patients (67% women and 33% men) who underwent 99mTc-sestamibi MPI were collected. The images were manually labeled with normal and abnormal labels according to the doctor’s diagnosis reports. The proposed deep learning model was trained using stress and rest polar maps and evaluated for prediction of obstructive disease in a stratified 5-fold cross-validation procedure.Results: The mean values of accuracy, sensitivity, accuracy, specificity, f1 score, and the area under the roc curve were 0.7562, 0.7856, 0.5748, 0.7434, 0.6646, and, 0.8450, respectively over 5 folds using both stress and rest scans. The inclusion of rest perfusion maps significantly improved AUC of the deep learning model (AUC: 0.845; 95% CI: 0.832-0.857), compared with using stress polar maps only (AUC: 0.827; 95% CI: 0.814-0.840); P < 0.05.Conclusion: The results of the present work reveal the possible applications of deep learning for polar map images classification in SPECT MPI.

An assessment of attenuation correction of SPECT MPI images generated by deep learning model

This article evaluates the effectiveness of using a deep learning network model to generate reliable attenuation corrected the single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI). The authors collected myocardial perfusion imaging data of 88 patients from a SPECT/CT machine, with an average age of 62.47 years. Then, two datasets are created from the original data: set A includes the deep learning-based attenuation corrected images (Generated Attenuation Correction - GenAC), and the non-attenuation corrected images; set B contains only non-attenuation corrected images. These datasets were diagnosed by two doctors (in which, one has 7 years of experience and the other has 10 years of experience in reading SPECT MPI). The doctors diagnose based on the image data without knowing which dataset it belongs to. The sensitivity, specificity, diagnostic accuracy, and lesion rate were evaluated between the two data sets. Results: The average specificity, sensitivity, and accuracy of the set with the deep learning-based attenuation corrected images were 0.87, 0.86, 0.86, while the results with the non-attenuation corrected images are 0.69, 0.83, and 0.78.