Table 2.
In vivo classification performance for different machine learning models, feature selection strategies, and VMI energies.
| Model | Features | Energy (keV) | AUC | Brier Score (BS) | R2 (Scaled BS) |
|---|---|---|---|---|---|
| Penalized Logistic Regression | all | 40 | 0.84 (0.95; 0.72) | 0.45 (0.29; 0.69) | 0.29 (−0.11; 0.54) |
| 50 | 0.82 (0.93; 0.69) | 0.48 (0.32; 0.73) | 0.25 (−0.15; 0.49) | ||
| 75 | 0.80 (0.92; 0.70) | 0.50 (0.33; 0.76) | 0.22 (−0.18; 0.48) | ||
| 120 | 0.85 (0.94; 0.75) | 0.43 (0.31; 0.57) | 0.32 (0.11; 0.50) | ||
| 190 | 0.84 (0.94; 0.71) | 0.45 (0.32; 0.63) | 0.29 (0.026; 0.49) | ||
| repeatable | 40 | 0.81 (0.93; 0.68) | 0.50 (0.34; 0.72) | 0.22 (−0.13; 0.46) | |
| 50 | 0.84 (0.94; 0.72) | 0.45 (0.27; 0.65) | 0.29 (−0.0076; 0.57) | ||
| 75 | 0.81 (0.92; 0.69) | 0.50 (0.33; 0.69) | 0.22 (−0.091; 0.48) | ||
| 120 | 0.79 (0.91; 0.67) | 0.51 (0.36; 0.70) | 0.21 (−0.095; 0.43) | ||
| 190 | 0.80 (0.91; 0.66) | 0.51 (0.36; 0.70) | 0.20 (−0.079; 0.42) | ||
| non-repeatable | 40 | 0.84 (0.96; 0.72) | 0.45 (0.28; 0.71) | 0.29 (−0.13; 0.55) | |
| 50 | 0.81 (0.93; 0.68) | 0.48 (0.33; 0.68) | 0.25 (−0.063; 0.48) | ||
| 75 | 0.80 (0.92; 0.69) | 0.50 (0.34; 0.74) | 0.22 (−0.15; 0.46) | ||
| 120 | 0.84 (0.94; 0.73) | 0.45 (0.33; 0.57) | 0.30 (0.11; 0.48) | ||
| 190 | 0.83 (0.93; 0.70) | 0.47 (0.34; 0.65) | 0.27 (−0.0039; 0.46) | ||
| Random Forest | all | 40 | 0.84 (0.94; 0.72) | 0.45 (0.33; 0.58) | 0.30 (0.086; 0.47) |
| 50 | 0.85 (0.94; 0.74) | 0.45 (0.34; 0.59) | 0.30 (0.089; 0.47) | ||
| 75 | 0.81 (0.92; 0.69) | 0.48 (0.36; 0.62) | 0.25 (0.04; 0.43) | ||
| 120 | 0.80 (0.92; 0.67) | 0.49 (0.37; 0.63) | 0.23 (0.026; 0.41) | ||
| 190 | 0.79 (0.91; 0.66) | 0.50 (0.38; 0.64) | 0.21 (0.016; 0.38) | ||
| repeatable | 40 | 0.83 (0.93; 0.72) | 0.47 (0.35; 0.61) | 0.27 (0.038; 0.46) | |
| 50 | 0.82 (0.93; 0.71) | 0.48 (0.36; 0.62) | 0.25 (0.016; 0.43) | ||
| 75 | 0.80 (0.91; 0.68) | 0.50 (0.37; 0.64) | 0.22 (−0.0043; 0.41) | ||
| 120 | 0.79 (0.90; 0.66) | 0.51 (0.37; 0.65) | 0.20 (−0.0044; 0.39) | ||
| 190 | 0.77 (0.88; 0.64) | 0.53 (0.41; 0.67) | 0.17 (−0.037; 0.36) | ||
| non-repeatable | 40 | 0.85 (0.94; 0.74) | 0.46 (0.35; 0.57) | 0.29 (0.092; 0.44) | |
| 50 | 0.86 (0.94; 0.75) | 0.44 (0.34; 0.56) | 0.31 (0.11; 0.46) | ||
| 75 | 0.82 (0.92; 0.72) | 0.47 (0.36; 0.59) | 0.26 (0.08; 0.42) | ||
| 120 | 0.81 (0.92; 0.68) | 0.49 (0.38; 0.62) | 0.24 (0.051; 0.4) | ||
| 190 | 0.81 (0.92; 0.68) | 0.49 (0.38; 0.61) | 0.23 (0.05; 0.39) |
AUC = area under the receiver operating curve.