Article Figures & Data
Figures
- FIGURE 1.
Outline of model architecture. ConvLSTM includes network trained to segment CAC, as well as second network for segmentation of heart, which limits CAC scoring. Softmax argmax function normalizes output of network to expected probabilities. Model identifies coronary calcium (red) and noncoronary calcium (green) within heart mask.
- FIGURE 2.
Examples of expert scores compared with DL CAC scores. Model identifies coronary calcium (red) and noncoronary calcium (green). In case 1, expert and DL annotations identified similar left circumflex CAC as well as ascending aorta calcium. No CAC was identified by either expert or DL scoring in case 2. In case 3, expert and DL annotations identified similar right coronary artery CAC as well as mitral annular calcification. BMI = body mass index.
- FIGURE 3.
Concordance matrix between DL and expert CAC categories in external testing population.
- FIGURE 4.
Kaplan–Meier survival curves for MACE. Increasing CAC category was associated with increasing risk of MACE for DL and expert annotated CAC scores on SPECT/CT attenuation maps.
- FIGURE 5.
Results of net-reclassification analysis. We assessed addition of CAC categories to full multivariable model outlined in Table 2.
Tables
- TABLE 1.
External Testing: Patient Characteristics According to CAC Category Determined by Deep-Learning Model
Characteristic CAC < 1 CAC 1–100 CAC 100–400 CAC > 400 P n 908 (40.0%) 596 (26.2%) 354 (15.6%) 413 (18.2%) Age (y) 61.9 (55.1–69.3) 66.4 (57.3–74.2) 70.8 (65.3–77.3) 72.3 (66.3–77.9) <0.001 Male 368 (40.5%) 293 (49.2%) 200 (56.5%) 286 (69.2%) <0.001 BMI 29.3 (25.1–32.6) 30 (25.8–34.4) 29.3 (25.4–32.9) 29.4 (25.2–32.4) 0.048 Past medical history Hypertension 423 (46.6%) 355 (59.6%) 240 (67.8%) 268 (64.9%) <0.001 Diabetes 136 (15.0%) 146 (24.5%) 111 (31.4%) 140 (33.9%) <0.001 Dyslipidemia 334 (36.8%) 246 (41.3%) 187 (52.8%) 236 (57.1%) <0.001 Family history 453 (49.9%) 305 (51.2%) 155 (43.8%) 205 (49.6%) 0.20 Smoking 67 (7.4%) 35 (5.9%) 21 (5.9%) 27 (6.5%) 0.67 BMI = body mass index.
Qualitative data are number and percentage; continuous data are median and interquartile range.
Association Unadjusted HR Adjusted HR 95% CI P 95% CI P DL CAC categories <1 Reference — Reference — 1–100 2.20 (1.54–3.14) <0.001 1.90 (1.32–2.73) <0.001 101–400 4.58 (3.23–6.48) <0.001 3.32 (2.29–4.81) <0.001 >400 5.92 (4.27–8.22) <0.001 3.58 (2.47–5.19) <0.001 Age (per 10 y) 1.37 (1.24–1.52) <0.001 1.12 (1.00–1.26) 0.046 Male 1.75 (1.39–2.19) <0.001 1.11 (0.86–1.43) 0.418 BMI (per kg/m2) 0.98 (0.96–1.00) 0.021 0.99 (0.97–1.01) 0.157 Hypertension 1.22 (0.98–1.53) 0.079 0.98 (0.77–1.25) 0.862 Diabetes 1.60 (1.26–2.02) <0.001 1.28 (0.99–1.64) 0.060 Dyslipidemia 1.34 (1.08–1.67) 0.008 1.00 (0.78–1.27) 0.997 Family history 0.82 (0.65–1.02) 0.071 0.90 (0.72–1.13) 0.353 Smoking 1.18 (0.81–1.72) 0.389 1.18 (0.80–1.74) 0.415 Stress AC TPD category < 1% Reference — Reference — 1–<5% 1.28 (0.96–1.71) 0.097 1.22 (0.90–1.65) 0.200 5–<10% 2.06 (1.46–2.90) <0.001 1.70 (1.19–2.44) 0.004 ≥10% 7.52 (5.43–10.4) <0.001 4.73 (3.02–7.46) <0.001 Rest AC TPD 1.07 (1.05–1.08) <0.001 1.00 (0.97–1.03) 0.836 Stress LVEF 0.97 (0.97–0.98) <0.001 0.99 (0.98–1.00) 0.293 BMI = body mass index; AC = attenuation correction; TPD = total perfusion deficit; LVEF = left ventricular ejection fraction.
Supplemental Data
Files in this Data Supplement:
