Table 6.
ICC values of subregion volume comparison between multi-modality model with certain modalities missing and model without modality missing
| Model | Missing modality | CA1 | CA2 | CA3 | DG | SUB | BA35 | BA36 | ERC | PHC |
|---|---|---|---|---|---|---|---|---|---|---|
|
| ||||||||||
| Multi-modality Model with ModAug | 3T | 0.995 | 0.989 | 0.981 | 0.972 | 0.992 | 0.971 | 0.98 | 0.988 | 0.986 |
| 7T-T1w | 0.998 | 0.994 | 0.99 | 0.997 | 0.998 | 0.986 | 0.998 | 0.993 | 0.992 | |
| 3T and 7T-T1w | 0.871 | 0.811 | 0.793 | 0.869 | 0.854 | 0.758 | 0.767 | 0.851 | 0.804 | |
|
| ||||||||||
| Single-modality Model | 3T and 7T-T1w * | 0.795 | 0.747 | 0.662 | 0.783 | 0.75 | 0.696 | 0.675 | 0.595 | 0.786 |
*
The single-modality model only used 7T-T2w as input. Therefore, there was no 3T or 7T-T1w fed into the model. We also call this situation as “missing 3T and 7T-T1w” here.