Interpretation of SPECT wall motion with deep learning.

OBJECTIVES: We sought to develop a novel deep learning (DL) workflow to interpret single-photon emission computed tomography (SPECT) wall motion. BACKGROUND: Wall motion assessment with SPECT is limited by image temporal and spatial resolution. Visual interpretation of wall motion can be subjective and prone to error. Artificial intelligence (AI) may improve accuracy of wall motion assessment. METHODS: A total of 1038 patients undergoing rest electrocardiogram (ECG)-gated SPECT and echocardiography were included. Using echocardiography as truth, a DL-model (DL-model 1) was trained to predict the probability of abnormal wall motion. Of the 1038 patients, 317 were used to train a DL-model (DL-model 2) to assess regional wall motion. A 10-fold cross-validation was adopted. Diagnostic performance of DL was compared with human readers and quantitative parameters. RESULTS: The area under the receiver operating characteristic curve (AUC) and accuracy (ACC) of DL model (AUC: .82 [95% CI: .79-.85]; ACC: .88) were higher than human (AUC: .77 [95% CI: .73-.81]; ACC: .82; P < .001) and quantitative parameter (AUC: .74 [95% CI: .66-.81]; ACC: .78; P < .05). The net reclassification index (NRI) was 7.7%. The AUC and accuracy of DL model for per-segment and per-vessel territory diagnosis were also higher than human reader. The DL model generated results within 30 seconds with operable guided user interface (GUI) and therefore could provide preliminary interpretation. CONCLUSIONS: DL can be used to improve interpretation of rest SPECT wall motion as compared with current human readers and quantitative parameter diagnosis.

Utility of FFRCT in Patients with Chest Pain.

PURPOSE OF REVIEW: The goal of this article is to review the data supporting the use of fractional flow reserve derived from coronary computed tomography angiography (FFRCT) in patients with chest pain. REVIEW FINDINGS: Numerous clinical trials have demonstrated that the diagnostic accuracy of coronary computed tomography angiography (CCTA) can be improved with the use of FFRCT, primarily due to its superior specificity when compared to CCTA alone. This promising development may help reduce the need for invasive angiography in patients presenting with chest pain. Furthermore, some studies have indicated that incorporating FFRCT into decision-making is safe, with an FFRCT value of ≥ 0.8 being associated with favorable outcomes. While FFRCT has been shown to be feasible in patients with acute chest pain, further large-scale studies are warranted to confirm its utility. The emergence of FFRCT as a tool for the management of patients with chest pain is promising. However, potential limitations require the interpretation of FFRCT in conjunction with clinical context.

Evaluation of an artificial intelligence coronary artery calcium scoring model from computed tomography.

OBJECTIVES: Coronary artery calcium (CAC) scores derived from computed tomography (CT) scans are used for cardiovascular risk stratification. Artificial intelligence (AI) can assist in CAC quantification and potentially reduce the time required for human analysis. This study aimed to develop and evaluate a fully automated model that identifies and quantifies CAC. METHODS: Fully convolutional neural networks for automated CAC scoring were developed and trained on 2439 cardiac CT scans and validated using 771 scans. The model was tested on an independent set of 1849 cardiac CT scans. Agatston CAC scores were further categorised into five risk categories (0, 1-10, 11-100, 101-400, and > 400). Automated scores were compared to the manual reference standard (level 3 expert readers). RESULTS: Of 1849 scans used for model testing (mean age 55.7 ± 10.5 years, 49% males), the automated model detected the presence of CAC in 867 (47%) scans compared with 815 (44%) by human readers (p = 0.09). CAC scores from the model correlated very strongly with the manual score (Spearman's r = 0.90, 95% confidence interval [CI] 0.89-0.91, p < 0.001 and intraclass correlation coefficient = 0.98, 95% CI 0.98-0.99, p < 0.001). The model classified 1646 (89%) into the same risk category as human observers. The Bland-Altman analysis demonstrated little difference (1.69, 95% limits of agreement: -41.22, 44.60) and there was almost excellent agreement (Cohen's κ = 0.90, 95% CI 0.88-0.91, p < 0.001). Model analysis time was 13.1 ± 3.2 s/scan. CONCLUSIONS: This artificial intelligence-based fully automated CAC scoring model shows high accuracy and low analysis times. Its potential to optimise clinical workflow efficiency and patient outcomes requires evaluation. KEY POINTS: • Coronary artery calcium (CAC) scores are traditionally assessed using cardiac computed tomography and require manual input by human operators to identify calcified lesions. • A novel artificial intelligence (AI)-based model for fully automated CAC scoring was developed and tested on an independent dataset of computed tomography scans, showing very high levels of correlation and agreement with manual measurements as a reference standard. • AI has the potential to assist in the identification and quantification of CAC, thereby reducing the time required for human analysis.

Prognostic Significance of Nonobstructive Left Main Coronary Artery Disease in Patients With and Without Diabetes: Long-Term Outcomes From the CONFIRM Registry.

BACKGROUND: Prognostic significance of non-obstructive left main (LM) disease was recently reported. However, the influence of diabetes mellitus (DM) on event rates in patients with and without non-obstructive LM disease is not well-known. METHODS: We evaluated 27,252 patients undergoing coronary computed tomographic angiography from the COroNary CT Angiography Evaluation For Clinical Outcomes: An InteRnational Multicenter (CONFIRM) Registry. Cumulative long-term incidence of all-cause mortality (ACM) was assessed between DM and non-DM patients by normal or non-obstructive LM disease (1-49% stenosis). RESULTS: The mean age of the study population was 57.6±12.6 years. Of the 27,252 patients, 4,434 (16%) patients had DM. A total of 899 (3%) deaths occurred during the follow-up of 3.6±1.9. years. Compared to patients with normal LM, those with non-obstructive LM had more pronounced overall coronary atherosclerosis and more cardiovascular risk factors. After clinical risk factors, segment involvement score, and stenosis severity adjustment, compared to patients without DM and normal LM, patients with DM were associated with increased ACM regardless of normal (HR 1.48, 95% CI 1.22-1.78, p<0.001) or non-obstructive LM (HR 1.46, 95% CI 1.04-2.04, p=0.029), while nonobstructive LM disease was not associated with increased ACM in patients without DM (HR 0.85, 95% CI 0.67-1.07, p=0.165) and there was no significant interaction between DM and LM status (HR 1.03, 95% CI 0.69-1.54, p=0.879). CONCLUSION: From the CONFIRM registry, we demonstrated that DM was associated with increased ACM. However, the presence of non-obstructive LM was not an independent risk marker of ACM, and there was no significant interaction between DM and non-obstructive LM disease for ACM.

OUTSMART HF: A Randomized Controlled Trial of Routine Versus Selective Cardiac Magnetic Resonance for Patients With Nonischemic Heart Failure (IMAGE-HF 1B).

BACKGROUND: Cardiac magnetic resonance (CMR) is a recommended imaging test for patients with heart failure (HF); however, there is a lack of evidence showing incremental benefit over transthoracic echocardiography. Our primary hypothesis was that routine use of CMR will yield more specific diagnoses in nonischemic HF. Our secondary hypothesis was that routine use of CMR will improve patient outcomes. METHODS: Patients with nonischemic HF were randomized to routine versus selective CMR. Patients in the routine strategy underwent echocardiography and CMR, whereas those assigned to selective use underwent echocardiography with or without CMR according to the clinical presentation. HF causes was classified from the imaging data as well as by the treating physician at 3 months (primary outcome). Clinical events were collected for 12 months. RESULTS: A total of 500 patients (344 male) with mean age 59±13 years were randomized. The routine and selective CMR strategies had similar rates of specific HF causes at 3 months clinical follow-up (44% versus 50%, respectively; P=0.22). At image interpretation, rates of specific HF causes were also not different between routine and selective CMR (34% versus 30%, respectively; P=0.34). However, 24% of patients in the selective group underwent a nonprotocol CMR. Patients with specific HF causes had more clinical events than those with nonspecific caused on the basis of imaging classification (19% versus 12%, respectively; P=0.02), but not on clinical assessment (15% versus 14%, respectively; P=0.49). CONCLUSIONS: In patients with nonischemic HF, routine CMR does not yield more specific HF causes on clinical assessment. Patients with specific HF causes from imaging had worse outcomes, whereas HF causes defined clinically did not. Registration: URL: https://www.clinicaltrials.gov; Unique identifier: NCT01281384.

Machine learning of clinical variables and coronary artery calcium scoring for the prediction of obstructive coronary artery disease on coronary computed tomography angiography: analysis from the CONFIRM registry.

AIMS: Symptom-based pretest probability scores that estimate the likelihood of obstructive coronary artery disease (CAD) in stable chest pain have moderate accuracy. We sought to develop a machine learning (ML) model, utilizing clinical factors and the coronary artery calcium score (CACS), to predict the presence of obstructive CAD on coronary computed tomography angiography (CCTA). METHODS AND RESULTS: The study screened 35 281 participants enrolled in the CONFIRM registry, who underwent ≥64 detector row CCTA evaluation because of either suspected or previously established CAD. A boosted ensemble algorithm (XGBoost) was used, with data split into a training set (80%) on which 10-fold cross-validation was done and a test set (20%). Performance was assessed of the (1) ML model (using 25 clinical and demographic features), (2) ML + CACS, (3) CAD consortium clinical score, (4) CAD consortium clinical score + CACS, and (5) updated Diamond-Forrester (UDF) score. The study population comprised of 13 054 patients, of whom 2380 (18.2%) had obstructive CAD (≥50% stenosis). Machine learning with CACS produced the best performance [area under the curve (AUC) of 0.881] compared with ML alone (AUC of 0.773), CAD consortium clinical score (AUC of 0.734), and with CACS (AUC of 0.866) and UDF (AUC of 0.682), P < 0.05 for all comparisons. CACS, age, and gender were the highest ranking features. CONCLUSION: A ML model incorporating clinical features in addition to CACS can accurately estimate the pretest likelihood of obstructive CAD on CCTA. In clinical practice, the utilization of such an approach could improve risk stratification and help guide downstream management.

Complementary pre-operative risk assessment using coronary computed tomography angiography and nuclear myocardial perfusion imaging in non-cardiac surgery: A VISION-CTA sub-study.

BACKGROUND: The incremental value and optimal utilization of non-invasive testing for prediction of peri-operative cardiac events during non-cardiac surgery are not clear. METHODS: A sub-study of VISION-CTA was performed using patients who underwent both coronary computed tomography angiography (CCTA) and nuclear myocardial perfusion imaging (MPI) as part of their pre-operative assessment. CCTA images were compared with MPI to determine the correlation between ischemia and obstructive coronary artery disease (CAD). Patients were followed post-operatively for 30 days and primary outcomes were all-cause death and non-fatal myocardial infarction. The predictive capacity of CCTA and nuclear MPI in predicting peri-operative major adverse cardiac event (MACE) was analyzed. RESULTS: A total of 55 patients (mean age 68.5 ± 8.4 years, 80.0% male) were analyzed. There was a strong correlation between the degree of obstructive CAD and the severity of perfusion abnormalities. Patients with severe CAD (≥ 70% stenosis) had a higher summed stress score than those without severe CAD [4.88 ± 1.22 and 1.30 ± 0.62, respectively (P < .05)]. Similarly summed difference score was significantly higher in patients with severe CAD [1.33 ± 0.46 and 0.17 ± 0.17 (P < .05)]. At 30 days there was a total of 8 (14.5%) MACE. The rate of MACE was higher in patients with severe CAD than those without (20.7% and 7.7%, respectively). Myocardial ischemia appeared to be predictive of MACE with an unadjusted odds ratio of 14.63 (P = .003). The predictive capacity of MPI further improved when only those patients with severe CAD were included (33.00) with a sensitivity, specificity, positive predictive value, and negative predictive value of 100% (79.4-100.0), 72.7% (49.8-89.3), 50.0% (21.1-78.9), and 100% (79.4-100.0), respectively. CONCLUSION: Although patients with significant obstructive disease are at risk of peri-operative MACE, the absolute event rate is low. Our data, albeit hypothesis generating, suggest that the peri-operative risk may be refined further by employing nuclear MPI in those with obstructive disease on CCTA.

Effect of proton pump inhibitors on Rubidium-82 gastric uptake using positron emission tomography myocardial perfusion imaging.

BACKGROUND: Rb-82 positron emission tomography (PET) myocardial perfusion imaging (MPI) is a robust tool for the evaluation of coronary artery disease (CAD). However, gastric uptake and spillover can be seen in 10% of Rb-82 PET MPI studies, commonly affecting the inferior wall, and can preclude the accurate identification of myocardial ischemia. We sought to understand the relationship between Rb-82 gastric uptake and the use of proton pump inhibitors (PPI). METHODS: 600 consecutive patients who presented for a clinically indicated Rb-82 PET MPI study were prospectively enrolled. In addition to the clinical history, PPI use was ascertained (medication, dose, frequency and duration of use, and time of last dose). Patients were categorized as PPI and non-PPI users. Rb-82 uptake in the gastrium, myocardium, and liver were measured at rest. Absolute uptake values and gastric:hepatic ratios were compared in PPI and non-PPI users. RESULT: Of 600 enrolled patients, 181 (30.2%) patients were using PPI. The gastric Rb-82 uptake in PPI users was 23% higher than non-PPI users (146 ± 52 kBq/cc vs 119 ± 40 kBq/cc, respectively; P < 0.001). The resting gastric:hepatic Rb-82 uptake ratio was also 23% higher in PPI vs non-PPI users (2.7 ± 1.0 vs 2.2 ± 0.8, respectively; P < 0.001). CONCLUSION: The gastric uptake of Rb-82 appears to be greater in patients actively using PPI and may identify a group who might be at greater risk of non-diagnostic Rb-82 PET MPI.

Impact of trainee involvement on patient radiation exposure and contrast volumes during invasive cardiac procedures.

PURPOSE: The impact of cardiology fellows (CFs) and interventional cardiology fellows (ICFs) on patient radiation and contrast exposure during diagnostic coronary angiography and percutaneous coronary intervention is unknown. METHODS: Between 2011 and 2014, 16,175 cases were retrospectively assessed involving 27 CFs, 22 ICFs and 24 staff as primary operators. RESULTS: During diagnostic coronary angiography, ICFs administered the lowest radiation dose (5,648±5,523 cGy*cm2; 1.30 ± 1.27 mSv)-achieving 22% less radiation than the staff (6,889±4,294 cGy*cm2; 1.58 ± 0.99 mSv) and 36% less than CFs (7,700±6,751 cGy*cm2; 1.77 ± 1.55 mSv) (p<0.01). When adjusted for access site, CFs administered more radiation than either the ICFs or staff. However, differences between ICFs and staff were exclusively observed during transradial procedures (p<0.01). With regards to contrast administration, ICFs administered less contrast (126.3 ± 57.6 mL) than either CFs (130±52.4 mL) or staff (132.7±47.6 mL) (p<0.01)-again, a finding isolated to the transradial cohort. Of the 6,751 percutaneous coronary intervention cases, no significant differences existed between the ICFs or staff cardiologists in patient radiation exposure-but a CF as the primary operator resulted in an 18% increase in radiation exposure. Notably, contrast use was not different amongst the types of operators (p<0.05). CONCLUSION: In conclusion, having a cardiology fellow as primary operator during invasive cardiac procedures increases patient radiation exposure and minimally increases contrast administration. Strategies to minimize patient radiation exposure while maintaining trainee involvement should be evaluated.

Relationship between Perioperative Hypotension and Perioperative Cardiovascular Events in Patients with Coronary Artery Disease Undergoing Major Noncardiac Surgery.

BACKGROUND: Perioperative hypotension is associated with cardiovascular events in patients having noncardiac surgery. It is unknown if the severity of preexisting coronary artery disease determines susceptibility to the cardiovascular risks of perioperative hypotension. METHODS: In this retrospective exploratory analysis of a substudy of an international prospective blinded cohort study, 955 patients 45 yr of age or older with history or risk factors for coronary artery disease underwent coronary computed tomographic angiography before elective inpatient noncardiac surgery. The authors evaluated the potential interaction between angiographic findings and perioperative hypotension (defined as systolic blood pressure less than 90 mmHg for a total of 10 min or more during surgery or for any duration after surgery and for which intervention was initiated) on the composite outcome of time to myocardial infarction or cardiovascular death up to 30 days after surgery. Angiography assessors were blinded to study outcomes; patients, treating clinicians, and outcome assessors were blinded to angiography findings. RESULTS: Cardiovascular events (myocardial infarction or cardiovascular death within 30 days after surgery) occurred in 7.7% of patients (74/955), including in 2.7% (8/293) without obstructive coronary disease or hypotension compared to 6.7% (21/314) with obstructive coronary disease but no hypotension (hazard ratio, 2.51; 95% CI, 1.11 to 5.66; P = 0.027), 8.8% (14/159) in patients with hypotension but without obstructive coronary disease (hazard ratio, 3.85; 95% CI, 1.62 to 9.19; P = 0.002), and 16.4% (31/189) with obstructive coronary disease and hypotension (hazard ratio, 7.34; 95% CI, 3.37 to 15.96; P < 0.001). Hypotension was independently associated with cardiovascular events (hazard ratio, 3.17; 95% CI, 1.99 to 5.06; P < 0.001). This association remained in patients without obstructive disease and did not differ significantly across degrees of coronary disease (P value for interaction, 0.599). CONCLUSIONS: In patients having noncardiac surgery, perioperative hypotension was associated with cardiovascular events regardless of the degree of coronary artery disease on preoperative coronary computed tomographic angiography.

Prognostic utility of splenic response ratio in dipyridamole PET myocardial perfusion imaging.

BACKGROUND: Cardiac magnetic resonance perfusion studies with adenosine stress have shown that splenic response can identify patients with inadequate pharmacologic stress. We investigate the incremental prognostic impact of splenic response ratio (SRR) in patients with normal Rubidium (Rb)-82 PET myocardial perfusion imaging (MPI). METHODS: Consecutive patients undergoing dipyridamole Rb-82 PET MPI for the evaluation of coronary artery disease were screened. Spleen and liver Rb-82 activity was measured and the SRR was calculated: SRR = (Spleen stress/Liver stress)/(Spleen rest/Liver rest). Major adverse cardiac events (MACE) were determined at 1 year of follow-up in patients with normal summed stress score and normal summed difference score. RESULTS: Of the 839 patients screened, the spleen was visualized in 703 (84%) of scans. There was significantly higher MACE observed in splenic non-responders vs splenic responders in both the normal SSS (7.8% vs 2.9%, P = .027) and the normal SDS groups (7.4% vs 2.2%, P = .014). In multivariate analysis in patients with normal SDS, splenic response was a significant, independent predictor of MACE (HR 2.97, 95% CI 1.10 to 8.04, P = .033). CONCLUSIONS: SRR is a novel imaging metric to identify patients with sub-maximal vasodilator stress and an incremental prognostic marker in patients with normal SDS and SSS (Clinical Trial Registration: https://clinicaltrials.gov/ct2/show/NCT01128023).

Clinical performance of Rb-82 myocardial perfusion PET and Tc-99m-based SPECT in patients with extreme obesity.

BACKGROUND: We evaluated the performance of stress imaging with technetium-99m-labeled tetrofosmin single-photon emission computed tomography (SPECT) and rubidium-82 positron emission tomography (PET) in patients with extreme obesity, defined as body mass index ≥40 kg/m2. METHODS: We identified patients with extreme obesity who underwent angiography in our center and either stress SPECT or PET within the previous six months. Cohorts of patients with extreme obesity and a <5% pretest likelihood of CAD who underwent SPECT (N = 25) or PET (N = 25) were also included. RESULTS: In total, 108 patients who underwent SPECT (N = 57) or PET (N = 51) were identified. Scan interpretation was classified as definitely normal or abnormal in 83.3% of PET and 60.5% of SPECT scans, respectively (P < .01). PET demonstrated higher diagnostic accuracy and normalcy rate. PET was found to have higher specificity for the pooled cohort. Similar findings were observed using stenosis cut-offs of ≥50% and ≥70%. CONCLUSIONS: In patients with extreme obesity, PET enabled more definitive scan interpretation with less artifact compared to SPECT. PET provided higher diagnostic accuracy and specificity in the detection of obstructive coronary artery disease.

Prognostic value of coronary computed tomographic angiography findings in asymptomatic individuals: a 6-year follow-up from the prospective multicentre international CONFIRM study.

Aim: The long-term prognostic benefit of coronary computed tomographic angiography (CCTA) findings of coronary artery disease (CAD) in asymptomatic populations is unknown. Methods and results: From the prospective multicentre international CONFIRM long-term study, we evaluated asymptomatic subjects without known CAD who underwent both coronary artery calcium scoring (CACS) and CCTA (n = 1226). Coronary computed tomographic angiography findings included the severity of coronary artery stenosis, plaque composition, and coronary segment location. Using the C-statistic and likelihood ratio tests, we evaluated the incremental prognostic utility of CCTA findings over a base model that included a panel of traditional risk factors (RFs) as well as CACS to predict long-term all-cause mortality. During a mean follow-up of 5.9 ± 1.2 years, 78 deaths occurred. Compared with the traditional RF alone (C-statistic 0.64), CCTA findings including coronary stenosis severity, plaque composition, and coronary segment location demonstrated improved incremental prognostic utility beyond traditional RF alone (C-statistics range 0.71-0.73, all P < 0.05; incremental χ2 range 20.7-25.5, all P < 0.001). However, no added prognostic benefit was offered by CCTA findings when added to a base model containing both traditional RF and CACS (C-statistics P > 0.05, for all). Conclusions: Coronary computed tomographic angiography improved prognostication of 6-year all-cause mortality beyond a set of conventional RF alone, although, no further incremental value was offered by CCTA when CCTA findings were added to a model incorporating RF and CACS.

Applying Modern Virtual and Augmented Reality Technologies to Medical Images and Models.

Recent technological innovations have created new opportunities for the increased adoption of virtual reality (VR) and augmented reality (AR) applications in medicine. While medical applications of VR have historically seen greater adoption from patient-as-user applications, the new era of VR/AR technology has created the conditions for wider adoption of clinician-as-user applications. Historically, adoption to clinical use has been limited in part by the ability of the technology to achieve a sufficient quality of experience. This article reviews the definitions of virtual and augmented reality and briefly covers the history of their development. Currently available options for consumer-level virtual and augmented reality systems are presented, along with a discussion of technical considerations for their adoption in the clinical environment. Finally, a brief review of the literature of medical VR/AR applications is presented prior to introducing a comprehensive conceptual framework for the viewing and manipulation of medical images in virtual and augmented reality. Using this framework, we outline considerations for placing these methods directly into a radiology-based workflow and show how it can be applied to a variety of clinical scenarios.

Prognostic value of vasodilator response using rubidium-82 positron emission tomography myocardial perfusion imaging in patients with coronary artery disease.

BACKGROUND: Prognostic value of positron emission tomography (PET) myocardial perfusion imaging (MPI) is well established. There is paucity of data on how the prognostic value of PET relates to the hemodynamic response to vasodilator stress. We hypothesize that inadequate hemodynamic response will affect the prognostic value of PET MPI. METHODS AND RESULTS: Using a multicenter rubidium (Rb)-82 PET registry, 3406 patients who underwent a clinically indicated rest/stress PET MPI with a vasodilator agent were analyzed. Patients were categorized as, "responders" [increase in heart rate ≥ 10 beats per minute (bpm) and decrease in systolic blood pressure (SBP) ≥10 mmHg], "partial responders" (either a change in HR or SBP), and "non-responders" (no change in HR or SBP). Primary outcome was all-cause death (ACD), and secondary outcome was cardiac death (CD). Ischemic burden was measured using summed stress score (SSS) and % left ventricular (LV) ischemia. After a median follow-up of 1.68 years (interquartile range = 1.17- 2.55), there were 7.9% (n = 270) ACD and 2.6% (n = 54) CD. Responders with a normal PET MPI had an annualized event rate (AER) of 1.22% (SSS of 0-3) and 1.58% (% LV ischemia = 0). Partial and non-responders had higher AER with worsening levels of ischemic burden. In the presence of severe SSS ≥12 and LV ischemia of ≥10%, partial responders had an AER of 10.79% and 10.36%, compared to non-responders with an AER of 19.4% and 12.43%, respectively. Patient classification was improved when SSS was added to a model containing clinical variables (NRI: 42%, p < 0.001) and responder category was added (NRI: 61%, p < 0.001). The model including clinical variables, SSS and hemodynamic response has good discrimination ability (Harrell C statistics: 0.77 [0.74-0.80]). CONCLUSION: Hemodynamic response during a vasodilator Rb-82 PET MPI is predictive of ACD. Partial and non-responders may require additional risk stratification leading to altered patient management.

Prognostic value of age adjusted segment involvement score as measured by coronary computed tomography: a potential marker of vascular age.

Extent of coronary atherosclerotic disease (CAD) burden on coronary computed tomography angiography (CCTA) as measured by segment involvement score (SIS) has a prognostic value. We sought to investigate the incremental prognostic value of 'age adjusted SIS' (aSIS), which may be a marker of premature atherosclerosis and vascular age. Consecutive patients were prospectively enrolled into the CONFIRM (Coronary CT Angiography EvaluatioN For Clinical Outcomes: An InteRnational Multicentre) multinational observational study. Patients were followed for the outcome of all-cause death. aSIS was calculated on CCTA for each patient, and its incremental prognostic value was evaluated. A total of 22,211 patients [mean age 58.5 ± 12.7 years, 55.8% male) with a median follow-up of 27.3 months (IQR 17.8, 35.4)] were identified. After adjustment for clinical factors and presence of obstructive CAD, higher aSIS was associated with increased death on multivariable analysis, with hazard ratio (HR) 2.40 (1.83-3.16, p < 0.001), C-statistic 0.723 (0.700-0.756), net reclassification improvement (NRI) 0.36 (0.26-0.47, p < 0.001), and relative integrated discrimination improvement (IDI) 0.33 (p = 0.009). aSIS had HR 3.48 (2.33-5.18, p < 0.001) for mortality in those without obstructive CAD, compared to HR 1.79 (1.25-2.58, p = 0.02) in those with obstructive CAD. In conclusion, aSIS has an incremental prognostic value to traditional risk factors and obstructive CAD, and may enhance CCTA risk stratification.

Machine learning for prediction of all-cause mortality in patients with suspected coronary artery disease: a 5-year multicentre prospective registry analysis.

AIMS: Traditional prognostic risk assessment in patients undergoing non-invasive imaging is based upon a limited selection of clinical and imaging findings. Machine learning (ML) can consider a greater number and complexity of variables. Therefore, we investigated the feasibility and accuracy of ML to predict 5-year all-cause mortality (ACM) in patients undergoing coronary computed tomographic angiography (CCTA), and compared the performance to existing clinical or CCTA metrics. METHODS AND RESULTS: The analysis included 10 030 patients with suspected coronary artery disease and 5-year follow-up from the COronary CT Angiography EvaluatioN For Clinical Outcomes: An InteRnational Multicenter registry. All patients underwent CCTA as their standard of care. Twenty-five clinical and 44 CCTA parameters were evaluated, including segment stenosis score (SSS), segment involvement score (SIS), modified Duke index (DI), number of segments with non-calcified, mixed or calcified plaques, age, sex, gender, standard cardiovascular risk factors, and Framingham risk score (FRS). Machine learning involved automated feature selection by information gain ranking, model building with a boosted ensemble algorithm, and 10-fold stratified cross-validation. Seven hundred and forty-five patients died during 5-year follow-up. Machine learning exhibited a higher area-under-curve compared with the FRS or CCTA severity scores alone (SSS, SIS, DI) for predicting all-cause mortality (ML: 0.79 vs. FRS: 0.61, SSS: 0.64, SIS: 0.64, DI: 0.62; P< 0.001). CONCLUSIONS: Machine learning combining clinical and CCTA data was found to predict 5-year ACM significantly better than existing clinical or CCTA metrics alone.

Prognostic significance of blood pressure response during vasodilator stress Rb-82 positron emission tomography myocardial perfusion imaging.

BACKGROUND: A drop in blood pressure (BP) or blunted BP response is an established high-risk marker during exercise myocardial perfusion imaging (MPI); however, data are sparse regarding the prognostic value of BP response in patients undergoing vasodilator stress rubidium-82 (Rb-82) Positron Emission Tomography (PET) MPI. METHODS AND RESULTS: From the PET Prognosis Multicenter Registry, a cohort of 3413 patients underwent vasodilator stress Rb-82 PET MPI with dipyridamole or adenosine. We used multivariable Cox proportional hazard regression to analyze the association with mortality of four BP variables: stress minus rest systolic BP (∆SBP), stress minus rest diastolic BP (∆DBP), resting systolic BP (rSBP), and resting diastolic BP (rDBP). Covariates that had univariate P values <.10 were entered into the multivariable model. After median 1.7 years follow-up, 270 patients died. In univariate analyses, ∆SBP (P = .082), rSBP (P = .008), and rDBP (P < .001) were of potential prognostic value (P < .10), but ∆DBP was not (P = .96). After adjustment for other clinical and MPI variables, ∆SBP no longer independently predicted mortality (P = .082); only lower rSBP (P = .026) and lower rDBP (P = .045) remained independently prognostic. CONCLUSIONS: In patients undergoing vasodilator stress MPI, only lower resting BP is an independent predictor of mortality along with other clinical and MPI variables; BP response does not appear to add to risk stratification in these patients.

Appropriate Use Criteria for Cardiac Computed Tomography: Impact on Diagnostic Utility.

BACKGROUND: Appropriate Use Criteria (AUC) guidelines for cardiac computed tomography (CCT) were developed to limit testing to reasonable clinical settings. However, significant testing is still done for inappropriate indications. This study investigates the impact of AUC on evaluability of CCT to determine if inappropriate tests result in a greater proportion of nondiagnostic results. METHODS: Investigators reviewed the medical records of 2417 consecutive patients who underwent CCT at the University of Ottawa Heart Institute. We applied the 2010 AUC and classified them as appropriate, inappropriate, or uncertain. Unclassifiable tests, as well as those with uncertain appropriateness, were excluded from the final analysis. Cardiac computed tomography results were classified as diagnostic if (1) all coronary segments were visualized, evaluable, and without obstructive stenosis; or (2) obstructive coronary artery disease with greater than 50% diameter stenosis in at least 1 coronary artery. All other test results were considered nondiagnostic. RESULTS: Of the 1984 patients included in the final analysis, 1522 patients (76.7%) had indications that were appropriate, whereas the remaining 462 (23.3%) were inappropriate. Inappropriate tests resulted in a higher rate of nondiagnostic results compared with appropriate CCT (9.0% vs 6.2%, P = 0.034). Inappropriate tests also had significantly more studies with nonevaluable segments than appropriate tests (24.5% vs 16.4%, P < 0.001) and were more likely to reveal obstructive coronary disease than appropriate CCT (50.5% vs 32.7%, P < 0.001). CONCLUSIONS: Cardiac computed tomography done for inappropriate indications may be associated with lower diagnostic yield and could impact future downstream resource utilization and health care costs.