Table 2. Comparison of diagnostic performance pooled across two readers between routine MRI, 1.5-mm MRI without DLR, and 1.5-mm MRI + DLR.
| Performance parameter | Routine MRI | 1.5-mm MRI without DLR | 1.5-mm MRI + DLR | P * | P † | |
|---|---|---|---|---|---|---|
| Performance in the entire patients | ||||||
| Accuracy | 84.2 (77.7, 89.1) [197/234] | 73.1 (65.5, 79.5) [171/234] | 80.8 (74.0, 86.1) [189/234] | 0.371 | 0.010 | |
| Sensitivity | 72.1 (56.7, 83.5) [49/68] | 83.8 (71.6, 91.4) [57/68] | 91.2 (82.3, 95.8) [62/68] | < 0.001 | 0.064 | |
| Specificity | 89.2 (82.6, 93.4) [148/166] | 68.7 (59.1, 76.9) [114/166] | 76.5 (67.7, 83.5) [127/166] | 0.004 | 0.042 | |
| Sensitivity in subgroup of patients with TLE according to lesion location | ||||||
| Hippocampus | 75.0 (59.9, 85.8) [51/68] | 86.8 (72.9, 94.1) [59/68] | 92.7 (84.0, 96.8) [63/68] | 0.001 | 0.031 | |
| Temporal lobe | 89.7 (75.1, 96.2) [61/68] | 82.4 (69.1, 90.7) [56/68] | 92.7 (84.0, 96.8) [63/68] | 0.341 | 0.012 | |
Data are pooled results across the two readers in % with the 95% confidence interval in parentheses and raw numbers in brackets.
*P-value for comparison between routine MRI and 1.5-mm MRI + DLR, †P-value for comparison between 1.5-mm MRI without DLR and 1.5-mm MRI + DLR. P-value < 0.016 indicates statistically significant difference with Bonferroni correction.
DLR = deep learning-based reconstruction, TLE = temporal lobe epilepsy