Table 5. Discriminant function determining disease classification accuracy of multimodal MRI analyses differentiating dementia subtypes & control involving dimension reduced.
| Modalities | Model accuracy (%) | Features (#) | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | Youden’s index (J) |
|---|---|---|---|---|---|---|---|
| All modalities | 87.8 | 8 | 95.2 | 85.0 | 87.0 | 94.4 | 80.2 |
| ASL | 57.1 | 1 | 68.4 | 65.2 | 61.9 | 71.4 | 33.6 |
| MRS | 23.8 | 1 | 50.0 | 77.0 | 57.1 | 71.4 | 26.9 |
| T1 | 78.3 | 7 | 85.7 | 76.0 | 75.0 | 86.4 | 61.7 |
| DTI | 73.3 | 4 | 80.0 | 75.0 | 80.0 | 75.0 | 55.0 |
Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and total model accuracy (Model accuracy) from each discriminant function analysis assessing the ability of each of the MRI modalities to differentiate participants with Alzheimer’s disease from those with Parkinson’s disease from the healthy control group. An embedded hierarchical stepwise regression reduced the model to the fewest possible metrics (Features) necessary to differentiate the groups. MRI modalities included: arterial spin labelling (ASL), magnetic resonance spectroscopy (MRS), structural T1, and diffusion tensor imaging (DTI).