Table 2.
Associations between deep learning segmentation–derived left ventricular mass index and prevalent disease
| N events† | Odds ratio with covariate (95% CI) |
|||
|---|---|---|---|---|
| LVMI (per 1 SD) | LVH | LVH (90th percentile) | ||
| Hypertension | ||||
| InlineVF | 11,271 | 1.43 (1.39–1.47) | 2.30 (2.15–2.46) | 2.33 (2.17–2.50) |
| Regression | 11,271 | 1.27 (1.24–1.30) | 1.67 (1.38–2.01) | 1.64 (1.53–1.76) |
| Segmentation | 11,271 | 1.55 (1.51–1.59) | 2.76 (2.51–3.04) | 2.39 (2.23–2.57) |
| Atrial fibrillation | ||||
| InlineVF | 1053 | 0.99 (0.93–1.05) | 1.19 (0.99–1.44) | 1.27 (1.04–1.53) |
| Regression | 1053 | 1.00 (0.93–1.07) | 1.13 (0.59–1.93) | 0.99 (0.80–1.21) |
| Segmentation | 1053 | 1.13 (1.06–1.21) | 1.75 (1.37–2.20) | 1.61 (1.34–1.93) |
| Heart failure | ||||
| InlineVF | 241 | 1.45 (1.29–1.63) | 2.92 (2.16–3.89) | 3.02 (2.23–4.04) |
| Regression | 241 | 1.39 (1.23–1.57) | 3.94 (1.75–7.67) | 2.36 (1.71–3.20) |
| Segmentation | 241 | 1.71 (1.51–1.93) | 4.67 (3.28–6.49) | 3.73 (2.78–4.95) |
LVH = left ventricular hypertrophy; LVMI = left ventricular mass index.
†
Total N = 37,261 with available phenotypic data and cardiac magnetic resonance–derived left ventricular mass estimates obtained using each method.