Table 1 Descriptive statistics for the effective diameter distributions of each morphological class using high-resolution images (227 × 227 pixels) of individually segmented RBCs from the cleaned MH dataset classified by the deep CNN ensemble (n = 1,294,996).
From: Deep ensemble learning enables highly accurate classification of stored red blood cell morphology
Morphology Class | Mean ± SD [μm] | Mean 95% CI [μm] | SD 95% CI [μm] | Mean MAD [μm] | Median (IQR) [μm] | Median 95% CI [μm] | IQR 95% CI [μm] | Median MAD [μm] | Sample size |
|---|---|---|---|---|---|---|---|---|---|
D | 7.821 ± 0.429 | 7.819–7.824 | 0.427–0.431 | 0.342 | 7.816 (0.580) | 7.813–7.819 | 0.577–0.584 | 0.290 | 121,571 |
E1 | 7.800 ± 0.581 | 7.799–7.802 | 0.580–0.583 | 0.458 | 7.798 (0.753) | 7.795–7.800 | 0.750–0.756 | 0.377 | 368,292 |
E2 | 7.304 ± 0.567 | 7.301–7.307 | 0.565–0.570 | 0.451 | 7.318 (0.753) | 7.314–7.322 | 0.748–0.758 | 0.376 | 112,885 |
E3 | 6.433 ± 0.490 | 6.431–6.435 | 0.488–0.491 | 0.390 | 6.387 (0.655) | 6.384–6.390 | 0.651–0.659 | 0.324 | 167,345 |
SE | 5.963 ± 0.348 | 5.961–5.964 | 0.347–0.349 | 0.268 | 5.948 (0.433) | 5.947–5.950 | 0.431–0.435 | 0.216 | 273,332 |
S | 5.904 ± 0.292 | 5.903–5.905 | 0.291–0.293 | 0.231 | 5.905 (0.385) | 5.904–5.907 | 0.384–0.387 | 0.193 | 224,930 |
ST | 7.080 ± 0.522 | 7.074–7.087 | 0.518–0.527 | 0.415 | 7.084 (0.696) | 7.076–7.092 | 0.685–0.706 | 0.348 | 26,641 |
