Hybridizing machine learning in survival analysis of cardiac PET/CT imaging.

BACKGROUND: Machine Learning (ML) allows integration of the numerous variables delivered by cardiac PET/CT, while traditional survival analysis can provide explainable prognostic estimates from a restricted number of input variables. We implemented a hybrid ML-and-survival analysis of multimodal PET/CT data to identify patients who developed myocardial infarction (MI) or death in long-term follow up. METHODS: Data from 739 intermediate risk patients who underwent coronary CT and selectively stress 15O-water-PET perfusion were analyzed for the occurrence of MI and all-cause mortality. Images were evaluated segmentally for atherosclerosis and absolute myocardial perfusion through 75 variables that were integrated through ML into an ML-CCTA and an ML-PET score. These scores were then modeled along with clinical variables through Cox regression. This hybridized model was compared against an expert interpretation-based and a calcium score-based model. RESULTS: Compared with expert- and calcium score-based models, the hybridized ML-survival model showed the highest performance (CI .81 vs .71 and .64). The strongest predictor for outcomes was the ML-CCTA score. CONCLUSION: Prognostic modeling of PET/CT data for the long-term occurrence of adverse events may be improved through ML imaging score integration and subsequent traditional survival analysis with clinical variables. This hybridization of methods offers an alternative to traditional survival modeling of conventional expert image scoring and interpretation.

Severe Acute Thrombocytopenia After Treatment with Tirofiban: A Case Series Approach.

Glycoprotein IIb/IIIa inhibitors are an adjuvant therapy for the treatment of patients with acute coronary syndromes. The main adverse reactions are bleeding and thrombocytopenia in 1-2% of cases. A 66-year-old woman arrived at the emergency department with ST-elevation MI. The catheterisation lab was busy, so she received thrombolytic therapy. Coronary angiography revealed a 90% stenosis in the middle segment of the left anterior descending artery and Thrombolysis in MI 2 flow. Subsequent percutaneous coronary intervention showed abundant thrombus and a coronary dissection and it was necessary to insert five drug-eluting stents. Non-fractionated heparin and a tirofiban infusion were used. After the percutaneous coronary intervention, she developed severe thrombocytopenia, haematuria and gingivorrhagia, for which infusion of tirofiban was suspended. In follow-up, no major bleeding or subsequent haemorrhagic complications were identified. It is crucial to distinguish between heparin-induced thrombocytopenia and thrombocytopenia caused by other drugs. A high level of suspicion should be employed in these cases.

Synthetic Attenuation Correction Maps for SPECT Imaging Using Deep Learning: A Study on Myocardial Perfusion Imaging.

(1) Background: The CT-based attenuation correction of SPECT images is essential for obtaining accurate quantitative images in cardiovascular imaging. However, there are still many SPECT cameras without associated CT scanners throughout the world, especially in developing countries. Performing additional CT scans implies troublesome planning logistics and larger radiation doses for patients, making it a suboptimal solution. Deep learning (DL) offers a revolutionary way to generate complementary images for individual patients at a large scale. Hence, we aimed to generate linear attenuation coefficient maps from SPECT emission images reconstructed without attenuation correction using deep learning. (2) Methods: A total of 384 SPECT myocardial perfusion studies that used 99mTc-sestamibi were included. A DL model based on a 2D U-Net architecture was trained using information from 312 patients. The quality of the generated synthetic attenuation correction maps (ACMs) and reconstructed emission values were evaluated using three metrics and compared to standard-of-care data using Bland-Altman plots. Finally, a quantitative evaluation of myocardial uptake was performed, followed by a semi-quantitative evaluation of myocardial perfusion. (3) Results: In a test set of 66 test patients, the ACM quality metrics were MSSIM = 0.97 ± 0.001 and NMAE = 3.08 ± 1.26 (%), and the reconstructed emission quality metrics were MSSIM = 0.99 ± 0.003 and NMAE = 0.23 ± 0.13 (%). The 95% limits of agreement (LoAs) at the voxel level for reconstructed SPECT images were: [-9.04; 9.00]%, and for the segment level, they were [-11; 10]%. The 95% LoAs for the Summed Stress Score values between the images reconstructed were [-2.8, 3.0]. When global perfusion scores were assessed, only 2 out of 66 patients showed changes in perfusion categories. (4) Conclusion: Deep learning can generate accurate attenuation correction maps from non-attenuation-corrected cardiac SPECT images. These high-quality attenuation maps are suitable for attenuation correction in myocardial perfusion SPECT imaging and could obviate the need for additional imaging in standalone SPECT scanners.

Multi-task Deep Learning of Myocardial Blood Flow and Cardiovascular Risk Traits from PET Myocardial Perfusion Imaging.

BACKGROUND: Advanced cardiac imaging with positron emission tomography (PET) is a powerful tool for the evaluation of known or suspected cardiovascular disease. Deep learning (DL) offers the possibility to abstract highly complex patterns to optimize classification and prediction tasks. METHODS AND RESULTS: We utilized DL models with a multi-task learning approach to identify an impaired myocardial flow reserve (MFR <2.0 ml/g/min) as well as to classify cardiovascular risk traits (factors), namely sex, diabetes, arterial hypertension, dyslipidemia and smoking at the individual-patient level from PET myocardial perfusion polar maps using transfer learning. Performance was assessed on a hold-out test set through the area under receiver operating curve (AUC). DL achieved the highest AUC of 0.94 [0.87-0.98] in classifying an impaired MFR in reserve perfusion polar maps. Fine-tuned DL for the classification of cardiovascular risk factors yielded the highest performance in the identification of sex from stress polar maps (AUC = 0.81 [0.73, 0.88]). Identification of smoking achieved an AUC = 0.71 [0.58, 0.85] from the analysis of rest polar maps. The identification of dyslipidemia and arterial hypertension showed poor performance and was not statistically significant. CONCLUSION: Multi-task DL for the evaluation of quantitative PET myocardial perfusion polar maps is able to identify an impaired MFR as well as cardiovascular risk traits such as sex, smoking and possibly diabetes at the individual-patient level.

Myocardial perfusion quantification with Rb-82 PET: good interobserver agreement of Carimas software on global, regional, and segmental levels.

PURPOSE: To estimate the interobserver agreement of the Carimas software package (SP) on global, regional, and segmental levels for the most widely used myocardial perfusion PET tracer-Rb-82. MATERIALS AND METHODS: Rest and stress Rb-82 PET scans of 48 patients with suspected or known coronary artery disease (CAD) were analyzed in four centers using the Carimas SP. We considered values to agree if they simultaneously had an intraclass correlation coefficient (ICC) > 0.75 and a difference < 20% of the median across all observers. RESULTS: The median values on the segmental level were 1.08 mL/min/g for rest myocardial blood flow (MBF), 2.24 mL/min/g for stress MBF, and 2.17 for myocardial flow reserve (MFR). For the rest MBF and MFR, all the values at all the levels fulfilled were in excellent agreement. For stress MBF, at the global and regional levels, all the 24 comparisons showed excellent agreement. Only 1 out of 102 segmental comparisons (seg. 14) was over the adequate agreement limit-23.5% of the median value (ICC = 0.95). CONCLUSION: Interobserver agreement for Rb-82 PET myocardial perfusion quantification analyzed with Carimas is good at any LV segmentation level-global, regional, and segmental. It is good for all the estimates-rest MBF, stress MBF, and MFR.

Artificial Intelligence to Improve Risk Prediction with Nuclear Cardiac Studies.

PURPOSE OF REVIEW: As machine learning-based artificial intelligence (AI) continues to revolutionize the way in which we analyze data, the field of nuclear cardiology provides fertile ground for the implementation of these complex analytics. This review summarizes and discusses the principles regarding nuclear cardiology techniques and AI, and the current evidence regarding its performance and contribution to the improvement of risk prediction in cardiovascular disease. There is a growing body of evidence on the experimentation with and implementation of machine learning-based AI on nuclear cardiology studies both concerning SPECT and PET technology for the improvement of risk-of-disease (classification of disease) and risk-of-events (prediction of adverse events) estimations. These publications still report objective divergence in methods either utilizing statistical machine learning approaches or deep learning with varying architectures, dataset sizes, and performance. Recent efforts have been placed into bringing standardization and quality to the experimentation and application of machine learning-based AI in cardiovascular imaging to generate standards in data harmonization and analysis through AI. Machine learning-based AI offers the possibility to improve risk evaluation in cardiovascular disease through its implementation on cardiac nuclear studies. AI in improving risk evaluation in nuclear cardiology. * Based on the 2019 ESC guidelines.

Position paper of the EACVI and EANM on artificial intelligence applications in multimodality cardiovascular imaging using SPECT/CT, PET/CT, and cardiac CT.

In daily clinical practice, clinicians integrate available data to ascertain the diagnostic and prognostic probability of a disease or clinical outcome for their patients. For patients with suspected or known cardiovascular disease, several anatomical and functional imaging techniques are commonly performed to aid this endeavor, including coronary computed tomography angiography (CCTA) and nuclear cardiology imaging. Continuous improvement in positron emission tomography (PET), single-photon emission computed tomography (SPECT), and CT hardware and software has resulted in improved diagnostic performance and wide implementation of these imaging techniques in daily clinical practice. However, the human ability to interpret, quantify, and integrate these data sets is limited. The identification of novel markers and application of machine learning (ML) algorithms, including deep learning (DL) to cardiovascular imaging techniques will further improve diagnosis and prognostication for patients with cardiovascular diseases. The goal of this position paper of the European Association of Nuclear Medicine (EANM) and the European Association of Cardiovascular Imaging (EACVI) is to provide an overview of the general concepts behind modern machine learning-based artificial intelligence, highlights currently prefered methods, practices, and computational models, and proposes new strategies to support the clinical application of ML in the field of cardiovascular imaging using nuclear cardiology (hybrid) and CT techniques.

Improving patient identification for advanced cardiac imaging through machine learning-integration of clinical and coronary CT angiography data.

BACKGROUND: Standard computed tomography angiography (CTA) outputs a myriad of interrelated variables in the evaluation of suspected coronary artery disease (CAD). But an important proportion of obstructive lesions does not cause significant myocardial ischemia. Nowadays, machine learning (ML) allows integration of numerous variables through complex interdependencies that optimize classification and prediction at the individual level. We evaluated ML performance in integrating CTA and clinical variables to identify patients that demonstrate myocardial ischemia through PET and those who ultimately underwent early revascularization. METHODS AND RESULTS: 830 patients with CTA and selective PET were analyzed. Nine clinical and 58 CTA variables were integrated through ensemble-boosting ML to identify patients with ischemia and those who underwent early revascularization. ML performance was compared against expert CTA interpretation, calcium score and clinical variables. While ML using all CTA variables achieved an AUC = 0.85, it was outperformed by expert CTA interpretation (AUC = 0.87, p < 0.01 for comparison), comparable to ML integration of CTA variables with clinical variables. However, the best performance was achieved by ML integration of expert CTA interpretation and clinical variables for both dependent variables (AUCs = 0.91 and 0.90, p < 0.001). CONCLUSIONS: Machine learning integration of diagnostic CTA and clinical data may improve identification of patients with myocardial ischemia and those requiring early revascularization at the individual level. This could potentially aid in sparing the need for subsequent advanced imaging and better identifying patients in ultimate need for revascularization. While ML integrating all CTA variables did not outperform expert CTA interpretation, ML data integration from different sources consistently improves diagnostic performance.

Genome-Wide Association Study and Identification of a Protective Missense Variant on Lipoprotein(a) Concentration: Protective Missense Variant on Lipoprotein(a) Concentration-Brief Report.

[Figure: see text].

Validation of the European Society of Cardiology pre-test probability model for obstructive coronary artery disease.

AIMS: Estimation of pre-test probability (PTP) of disease in patients with suspected coronary artery disease (CAD) is a common challenge. Due to decreasing prevalence of obstructive CAD in patients referred for diagnostic testing, the European Society of Cardiology suggested a new PTP (2019-ESC-PTP) model. The aim of this study was to validate that model. METHODS AND RESULTS: Symptomatic patients referred for coronary computed tomography angiography (CTA) due to suspected CAD in a geographical uptake area of 3.3 million inhabitants were included. The reference standard was a combined endpoint of CTA and invasive coronary angiography (ICA) with obstructive CAD defined at ICA as a ≥50% diameter stenosis or fractional flow reserve ≤0.80 when performed. The 2019-ESC-PTP, 2013-ESC-PTP, and CAD Consortium basic PTP scores were calculated based on age, sex, and symptoms. Of the 42 328 identified patients, coronary stenosis was detected in 8.8% using the combined endpoint. The 2019-ESC-PTP and CAD Consortium basic scores classified substantially more patients into the low PTP groups (PTP < 15%) than did the 2013-ESC-PTP (64% and 65% vs. 16%, P < 0.001). Using the combined endpoint as reference, calibration of the 2019-ESC-PTP model was superior to the 2013-ESC-PTP and CAD Consortium basic score. CONCLUSION: The new 2019-ESC-PTP model is well calibrated and superior to the previously recommended models in predicting obstructive stenosis detected by a combined endpoint of CTA and ICA.

Sex-Based Differences in Unrecognized Myocardial Infarction.

Background Myocardial infarction is an important cause of morbidity and mortality in both men and women. Atypical or the absence of symptoms, more prevalent among women, may contribute to unrecognized myocardial infarctions and missed opportunities for preventive therapies. The aim of this research is to investigate sex-based differences of undiagnosed myocardial infarction in the general population. Methods and Results In the Lifelines Cohort Study, all individuals ≥18 years with a normal baseline ECG were followed from baseline visit till first follow-up visit (≈5 years, n=97 203). Individuals with infarct-related changes between baseline and follow-up ECGs were identified. The age- and sex-specific incidence rates were calculated and sex-specific cardiac symptoms and predictors of unrecognized myocardial infarction were determined. Follow-up ECG was available after a median of 3.8 (25th and 75th percentile: 3.0-4.6) years. During follow-up, 198 women experienced myocardial infarction (incidence rate 1.92 per 1000 persons-years) compared with 365 men (incidence rate 3.30; P<0.001 versus women). In 59 (30%) women, myocardial infarction was unrecognized compared with 60 (16%) men (P<0.001 versus women). Individuals with unrecognized myocardial infarction less often reported specific cardiac symptoms compared with individuals with recognized myocardial infarction. Predictors of unrecognized myocardial infarction were mainly hypertension, smoking, and higher blood glucose level. Conclusions A substantial proportion of myocardial infarctions are unrecognized, especially in women. Opportunities for secondary preventive therapies remain underutilized if myocardial infarction is unrecognized.

Integrating the STOP-BANG Score and Clinical Data to Predict Cardiovascular Events After Infarction: A Machine Learning Study.

BACKGROUND: OSA conveys worse clinical outcomes in patients with coronary artery disease. The STOP-BANG score is a simple tool that evaluates the risk of OSA and can be added to the large number of clinical variables and scores that are obtained during the management of patients with myocardial infarction (MI). Currently, machine learning (ML) is able to select and integrate numerous variables to optimize prediction tasks. RESEARCH QUESTION: Can the integration of STOP-BANG score with clinical data and scores through ML better identify patients who experienced an in-hospital cardiovascular event after acute MI? STUDY DESIGN AND METHODS: This is a prospective observational cohort study of 124 patients with acute MI of whom the STOP-BANG score classified 34 as low (27.4%), 30 as intermediate (24.2%), and 60 as high (48.4%) OSA-risk patients who were followed during hospitalization. ML implemented feature selection and integration across 47 variables (including STOP-BANG score, Killip class, GRACE score, and left ventricular ejection fraction) to identify those patients who experienced an in-hospital cardiovascular event (ie, death, ventricular arrhythmias, atrial fibrillation, recurrent angina, reinfarction, stroke, worsening heart failure, or cardiogenic shock) after definitive MI treatment. Receiver operating characteristic curves were used to compare ML performance against STOP-BANG score, Killip class, GRACE score, and left ventricular ejection fraction, independently. RESULTS: There were an increasing proportion of cardiovascular events across the low, intermediate, and high OSA risk groups (P = .005). ML selected 7 accessible variables (ie, Killip class, leukocytes, GRACE score, c reactive protein, oxygen saturation, STOP-BANG score, and N-terminal prohormone of B-type natriuretic peptide); their integration outperformed all comparators (area under the curve, 0.83 [95% CI, 0.74-0.90]; P < .01). INTERPRETATION: The integration of the STOP-BANG score into clinical evaluation (considering Killip class, GRACE score, and simple laboratory values) of subjects who were admitted for an acute MI because of ML can significantly optimize the identification of patients who will experience an in-hospital cardiovascular event.

Software reproducibility of myocardial blood flow and flow reserve quantification in ischemic heart disease: A 13N-ammonia PET study.

BACKGROUND: We explored agreement in the quantification of myocardial perfusion by cross-comparison of implemented software packages (SPs) in three distinguishable patient profile populations. METHODS: We studied 91 scans of patients divided into 3 subgroups based on their semi-quantitative perfusion findings: patients with normal perfusion, with reversible perfusion defects, and with fixed perfusion defects. Rest myocardial blood flow (MBF), stress MBF, and myocardial flow reserve (MFR) were obtained with QPET, SyngoMBF, and Carimas. Agreement between SPs was considered adequate when a pairwise standardized difference was found to be < 0.20 and its corresponding intraclass correlation coefficient was ≥ 0.75. RESULTS: In patients with normal perfusion, two out of three comparisons of global stress MBF quantifications were outside the limits of agreement. In ischemic patients, all comparisons of global stress MBF and MFR were outside the limits of established agreement. In patients with fixed perfusion defects, all SP comparisons of perfusion quantifications were within the limit of agreement. Regionally, agreement of these perfusion estimates was mostly found for the left anterior descending artery vascular territory. CONCLUSION: Reversible defects demonstrated the worst agreement in global stress MBF and MFR and discrepancies showed to be regional dependent. Reproducibility between SPs should not be assumed.

Ventricular synchrony is not significantly determined by absolute myocardial perfusion in patients with chronic heart failure: A 13N-ammonia PET study.

BACKGROUND: It is thought that heart failure (HF) patients may benefit from the evaluation of mechanical (dys)synchrony, and an independent inverse relationship between myocardial perfusion and ventricular synchrony has been suggested. We explore the relationship between quantitative myocardial perfusion and synchrony parameters when accounting for the presence and extent of fixed perfusion defects in patients with chronic HF. METHODS: We studied 98 patients with chronic HF who underwent rest and stress Nitrogen-13 ammonia PET. Multivariate analyses of covariance were performed to determine relevant predictors of synchrony (measured as bandwidth, standard deviation, and entropy). RESULTS: In our population, there were 43 (44%) women and 55 men with a mean age of 71 ± 9.6 years. The SRS was the strongest independent predictor of mechanical synchrony variables (p < .01), among other considered predictors including: age, sex, body mass index, smoking, diabetes mellitus, dyslipidemia, hypertension, rest myocardial blood flow (MBF), and myocardial perfusion reserve (MPR). Results were similar when considering stress MBF instead of MPR. CONCLUSIONS: The existence and extent of fixed perfusion defects, but not the quantitative PET myocardial perfusion parameters (sMBF and MPR), constitute a significant independent predictor of ventricular mechanical synchrony in patients with chronic HF.

Machine learning in the integration of simple variables for identifying patients with myocardial ischemia.

BACKGROUND: A significant number of variables are obtained when characterizing patients suspected with myocardial ischemia or at risk of MACE. Guidelines typically use a handful of them to support further workup or therapeutic decisions. However, it is likely that the numerous available predictors maintain intrinsic complex interrelations. Machine learning (ML) offers the possibility to elucidate complex patterns within data to optimize individual patient classification. We evaluated the feasibility and performance of ML in utilizing simple accessible clinical and functional variables for the identification of patients with ischemia or an elevated risk of MACE as determined through quantitative PET myocardial perfusion reserve (MPR). METHODS: 1,234 patients referred to Nitrogen-13 ammonia PET were analyzed. Demographic (4), clinical (8), and functional variables (9) were retrieved and input into a cross-validated ML workflow consisting of feature selection and modeling. Two PET-defined outcome variables were operationalized: (1) any myocardial ischemia (regional MPR < 2.0) and (2) an elevated risk of MACE (global MPR < 2.0). ROC curves were used to evaluate ML performance. RESULTS: 16 features were included for boosted ensemble ML. ML achieved an AUC of 0.72 and 0.71 in identifying patients with myocardial ischemia and with an elevated risk of MACE, respectively. ML performance was superior to logistic regression when the latter used the ESC guidelines risk models variables for both PET-defined labels (P < .001 and P = .01, respectively). CONCLUSIONS: ML is feasible and applicable in the evaluation and utilization of simple and accessible predictors for the identification of patients who will present myocardial ischemia and an elevated risk of MACE in quantitative PET imaging.