Variability in the accuracy of the exercise ST-segment in predicting the coronary angiogram: how good can we be?

Self-supervised deep representation learning of a foundation transformer model enabling efficient ECG-based assessment of cardiac and coronary function with limited labels

Background: Although deep learning methods have shown great promise for identification of structural and functional cardiac abnormalities using electrocardiographic data, these methods are data hungry, posing a challenge for critically important tasks where ground truth labels are relatively scarce. Impaired coronary microvascular and vasomotor function is difficult to identify with standard clinical methods of cardiovascular testing such as coronary angiography and noninvasive single photon emission tomography (SPECT) myocardial perfusion imaging (MPI). Gold standard data from positron emission tomography (PET) are gaining emphasis in clinical guidelines but are expensive and only available in relatively limited centers. We hypothesized that signals embedded within resting and stress electrocardiograms (ECGs) identify individuals with microvascular and vasomotor dysfunction. Methods: We developed and pretrained a self-supervised foundation vision transformer model using a large database of unlabeled ECG waveforms (N=800,035). We then fine-tuned the foundation model for two clinical tasks: the difficult problem of identifying patients with impaired myocardial flow reserve (AI-MFR), and the relatively easier problem of detecting impaired LVEF (AI-LVEF). A second ECG database was labeled with task-specific annotations derived from quantitative PET MPI (N=4167). Diagnostic accuracy of AI predictions was tested in a holdout set of patients undergoing PET MPI (N=1031). Prognostic evaluation was performed in the PET holdout cohort, as well as independent cohorts of patients undergoing pharmacologic or exercise stress SPECT MPI (N=6635). Results: The diagnostic accuracy of AI-MFR with SSL pretraining increased significantly compared to de novo supervised training (AUROC, sensitivity, specificity: 0.758, 70.1%, 69.4% vs. 0.632, 66.1%, 57.3%, p < 0.0001). SSL pretraining also produced a smaller increase in AI-LVEF accuracy (AUROC, sensitivity, specificity: 0.946, 89.4%, 85.9% vs. 0.918, 87.6%, 82.5%, p < 0.02). Abnormal AI-MFR was found to be significantly associated with mortality risk in all three test cohorts (Hazard Ratio (HR) 2.61 [95% CI 1.83, 3.71], p < 0.0001, PET cohort; HR 2.30 [2.03, 2.61], p < 0.0001, pharmacologic stress SPECT cohort; HR 3.76 [2.36, 5.99], p < 0.0001, exercise stress SPECT cohort). Conclusion: SSL pretraining of a vision transformer foundation model enabled identification of signals predictive of impaired MFR, a hallmark of microvascular and vasomotor dysfunction, and impaired LV function in resting and stress ECG waveforms. These signals are powerful predictors of prognosis in patients undergoing routine noninvasive stress testing and could enable more efficient diagnosis and management of these common conditions.

The ST segment depression pattern in asymptomatic peri-menopausal female athletes

Background

ST segment has not been well investigated in asymptomatic peri-menopausal female athletes, when the CV risk is higher.

Aims

The aims of the study is to investigate the prevalence of ST segment depression in peri-menopausal female athletes, divided in four age groups.

Methods

in a cohort of 6010 female athletes aged 45–65 years old, 161 subjects were selected for the presence of ST segment depression, revealed by maximal ergometric test. All athletes were also evaluated by physical examination and echocardiography. Inclusion criteria for ST segment depression were ST depression >0.5 mm respect to baseline and its depth was divided in 0.5 mm, 0.5–1 mm, 1–2 mm categories. Its behavior was classified in ascending, horizontal and descending and it was studied in relation to the age range (<51; 51–55; 56–60; >60).

Results

ST segment depression was mainly evident in inferolateral leads in all groups (63%, 70,3%, 71%, 63,6%, for <51; 51–55, 56–60 and 61–65 respectively) with horizontal pattern (52,2%, 59,5%, 57,8%, 63,6%) and 1–2 mm depth (43,5%, 46,7%, 60,6%), with the exception of the range 51–55, mainly showing a depth of 0,5-1 mm (45,9%). The older group showed increased SBP (137.5 mmHg, p = 0.007) and BMI (24.3, p = 0.093) values. Mitral prolapse was shown in 11,8% while 36,6% showed systolic flattening of mitral leaflets.

Conclusions

ST depression in asymptomatic menopausal female athletes is frequent and it is characterized by a specific presentation pattern. This is particularly important in the menopausal age when CV risk factors are more prevalent.

Detection of patients with coronary artery disease using cardiac magnetic field mapping at rest

We studied the use of cardiac magnetic field mapping to detect patients with CAD without subjecting them to stress. Fifty-nine healthy control subjects and 101 patients with CAD without previous MI were included. The optimal positions for detecting CAD were located in the left superior parasternal and in the inferior midsternal area. Values for ST slope, ST shift, T peak amplitude, ST-T integral, and magnetic field map orientation differed significantly between the 2 groups. Three parameters together in a multivariate analysis yielded a sensitivity of 84% and a specificity of 83% in distinguishing patients with CAD from control subjects. We suggest that cardiac magnetic field mapping is a promising technique to identify patients with CAD.

ST/HR hysteresis: exercise and recovery phase ST depression/heart rate analysis of the exercise ECG

ST segment depression/heart rate (ST/HR) hysteresis is a recently introduced novel computer method for integrating the exercise and recovery phase ST/HR analysis for improved detection of coronary artery disease (CAD). It is a continuous diagnostic variable that extracts the prevailing direction and average magnitude of the hysteresis in ST depression against HR during the first 3 consecutive minutes of postexercise recovery. This article reviews the development and evaluation of this new method in a clinical population of 347 patients referred for a routine bicycle exercise electrocardiographic (ECG) test at Tampere University Hospital, Finland. Of these patients, 127 had angiographically proven CAD, whereas 13 had no CAD according to angiography, 18 had no perfusion defect according to Tc-99m-sestamibi myocardial imaging and single photon emission computed tomography, and 189 were clinically normal with respect to cardiac diseases. For each patient, the values for ST/HR hysteresis, ST/HR index, end-exercise ST depression, and recovery ST depression were determined for each lead of the Mason-Likar modification of the standard 12-lead exercise ECG and maximum value from the lead system (aVL, aVR, and V1 excluded). The area under the receiver operating characteristics curve (ie, the discriminative capacity) of the ST/HR hysteresis was 89%, which was significantly larger than that of the end-exercise ST depression (76%, P < .0001), recovery ST depression (84%, P = .0063) or ST/HR index (83%, P = .0023), indicating the best diagnostic performance of the ST/HR hysteresis in detection of CAD regardless of the partition value selection. Furthermore, the superior diagnostic performance of the method was relatively insensitive to the ST segment measurement point or to the ECG lead selection. These results suggest that the ST/HR hysteresis improves the clinical utility of the exercise ECG test in detection of CAD.

Cardiac outcomes in coronary patients with submaximum dobutamine stress echocardiography

This study evaluated the prediction of cardiac events (cardiac death, myocardial infarction, unstable angina, or late myocardial revascularization) in patients with submaximum responses to dobutamine stress, defined by attainment of <85% age-predicted heart rate. Of 1,772 patients undergoing dobutamine echocardiography over a 2-year period, 425 had a submaximum heart rate response. After exclusion of patients treated with beta-adrenoceptor blocking agents, 255 patients formed the study group. In these patients, the test was terminated after administration of the maximum dose of 40 microg/kg/min of dobutamine with atropine (end of protocol, n = 186), severe angina, ischemic ST-segment changes, or intolerable side effects (n = 69). Dobutamine-induced changes (ischemia, viability, or both) were detected in 46 patients, involving ischemia in 133 segments, viability in 23, and ischemia and viability in 16 segments. Patients were followed for an interval of 28 +/- 17 months; 5 (1.2%) were lost to follow-up. Of the medically treated patients, cardiac events occurred in 73 of 228 (31%), including cardiac death in 25 (11%), nonfatal myocardial infarction in 11 (4.8%), severe unstable angina in 35 (15%), and late revascularization in 2 (0.9%). Cardiac events occurred in 11 of 30 (36%) with inducible abnormalities, and 62 of 198 without inducible abnormalities (31%, p = NS). Thus, cardiac event rates are high in patients with inadequate chronotropic responses to dobutamine stress, irrespective of whether stress-induced changes are detected. A negative dobutamine echocardiogram at submaximum heart rate should be considered nondiagnostic.

Accurate detection of coronary artery disease by integrated analysis of the ST-segment depression/heart rate patterns during the exercise and recovery phases of the exercise electrocardiography test

In this comparative cross-sectional study, we evaluated whether a novel computerized diagnostic variable, ST-segment depression/heart rate ST/HR analysis during both the exercise and postexercise recovery phases of the exercise electrocardiography (ECG) test, can detect coronary artery disease more accurately than methods using either exercise or recovery phase alone. The study population comprised 347 clinical patients referred for a routine bicycle exercise ECG test at Tampere University Hospital, Finland. Of these, 127 had angiographically proven coronary artery disease, whereas 13 had no coronary artery disease according to angiography, 18 had no perfusion defect according to technetium-99m sestamibi single-photon emission computed tomography, and 189 were clinically normal with respect to cardiac diseases. For each patient, the maximum values of the ST/HR hysteresis, ST/HR index, end-exercise ST depression, and recovery ST depression were determined from the Mason-Likar modification of the standard 12-lead exercise electrocardiogram [aVL, aVR, and V1 excluded]. The diagnostic performance of these continuous diagnostic variables was compared by means of receiver-operating characteristic analysis. The area under the receiver-operating characteristic curve of the ST/HR hysteresis was 89%, which was significantly larger than that of the end-exercise ST depression (76%, p < or = 0.0001), recovery ST depression (84%, p = 0.0063), or ST/HR index (83%, p = 0.0023), indicating superior diagnostic performance of the ST/HR hysteresis independent of the partition value selection. In conclusion, computerized analysis of the HR-adjusted ST depression pattern during the exercise phase, integrated with the HR-adjusted ST depression pattern during the recovery phase after exercise, can significantly improve the diagnostic performance and clinical utility of the exercise ECG test for the detection of coronary artery disease.

A computer program for comprehensive ST-segment depression/heart rate analysis of the exercise ECG test

The ST-segment depression/heart rate (ST/HR) analysis has been found to improve the diagnostic accuracy of the exercise ECG test in detecting myocardial ischemia. Recently, three different continuous diagnostic variables based on the ST/HR analysis have been introduced; the ST/HR slope, the ST/HR index and the ST/HR hysteresis. The latter utilises both the exercise and recovery phases of the exercise ECG test, whereas the two former are based on the exercise phase only. This present article presents a computer program which not only calculates the above three diagnostic variables but also plots the full diagrams of ST-segment depression against heart rate during both exercise and recovery phases for each ECG lead from given ST/HR data. The program can be used in the exercise ECG diagnosis of daily clinical practice provided that the ST/HR data from the ECG measurement system can be linked to the program. At present, the main purpose of the program is to provide clinical and medical researchers with a practical tool for comprehensive clinical evaluation and development of the ST/HR analysis.