Table 3.
Summary of APE distribution under different gamma criteria for ACLR and Poisson Lasso models in TV and CV.
| Metrics | 3%/3 mm |
3%/2 mm |
2%/2 mm |
||||
|---|---|---|---|---|---|---|---|
| ACLR, N (%) | PL, N (%) | ACLR, N (%) | PL, N (%) | ACLR, N (%) | PL, N (%) | ||
|
| |||||||
| TV | APE ⩽ 3.5% | 386 (90.6) | 369 (86.6) | 334 (78.4) | 304 (71.4) | 202 (47.4) | 186 (43.7) |
| APE ⩽ 5% | 405 (95.1) | 397 (93.2) | 375 (88.0) | 367 (86.2) | 265 (62.2) | 247 (58.0) | |
| APE ⩽ 10% | 419 (98.4) | 415 (97.4) | 416 (97.7) | 411 (96.5) | 385 (90.4) | 370 (86.9) | |
| MAE (SD) | 1.76% (1.8) | 2.10% (2.1) | 2.60 (2.4) | 3.04 (2.9) | 4.66% (3.9) | 5.28% (4.6) | |
| CV | APE ⩽ 3.5% | 135 (90.0) | 131 (87.3) | 113 (75.3) | 102 (68.0) | 39 (26.0) | 35 (23.3) |
| APE ⩽ 5% | 138 (92.0) | 133 (88.6) | 134 (89.3) | 132 (88.0) | 69 (46.0) | 53 (35.3) | |
| APE ⩽ 10% | 146 (95.3) | 143 (95.3) | 144 (96.0) | 142 (94.7) | 147 (98.0) | 141 (94) | |
| MAE (SD) | 1.73% (1.5) | 2.07% (1.7) | 2.99% (1.9) | 2.36% (2.1) | 5.93% (2.3) | 7.12% (3.2) | |
Abbreviations: APE = absolute prediction error; MAE = mean absolute error; SD = standard deviation; ACLR = autoencoder based classification-regression deep learning model; PL = Poisson Lasso; TV = technical validation