Table 1.
Assessment of image subsets via feature selection. Image subsets were selected by an exhaustive search using the Dice similarity coefficient (mean ± SD across all mice) between labels generated using the subset and the optimized protocol (i.e., T1 and T2 maps and all 22 saturation transfer-weighted images with B1 = 3 and 6 µT) as the metric. The positive and negative predictive value (PPV and NPV, respectively) of tumour and necrosis/apoptosis labels are also given with respect to those generated from all images from the optimized protocol.
| No. of Images | T1 | T2 | 3 µT | 6 µT | Dice similarity coefficient (%) |
Active tumour | Necrosis/apoptosis | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 3 ppm | 5 ppm | 30 ppm | 48 ppm | 5 ppm | 8 ppm | 48 ppm | 75 ppm | PPV (%) |
NPV (%) |
PPV (%) |
NPV (%) |
||||
| 3 | X | X | X | 93 ± 3 | 94 ± 4 | 96 ± 3 | 60 ± 30 | 99 ± 2 | |||||||
| 4 | X | X | X | X | 94 ±3 | 94 ± 5 | 97 ± 3 | 70 ± 30 | 98 ± 2 | ||||||
| 5 | X | X | X | X | X | 95 ± 2 | 97 ± 3 | 96 ± 3 | 70 ± 30 | 99 ± 1 | |||||
| 6 | X | X | X | X | X | X | 95 ± 2 | 97 ± 3 | 97 ± 3 | 70 ± 30 | 99 ± 1 | ||||
| 7 | X | X | X | X | X | X | X | 95 ± 2 | 95 ± 4 | 99 ± 1 | 70 ± 30 | 99 ± 2 | |||
| 8 | X | X | X | X | X | X | X | X | 97 ± 2 | 97 ± 3 | 99 ± 1 | 90 ± 10 | 99 ± 1 | ||
| 9 | X | X | X | X | X | X | X | X | X | 98 ± 1 | 98 ± 1 | 99 ± 1 | 95 ± 4 | 100 ± 1 | |