Table 2 Comparison of different fusion strategies; the numbers indicate out-of-fold means and standard deviation. The performance of the model is evaluated using the following metrics: balanced accuracy (BACC), sensitivity (SEN), specificity (SPE), and area under the curve (AUC).
From: Multimodal surface-based transformer model for early diagnosis of Alzheimer’s disease
Task | Dataset | Fusion | BACC | SEN | SPE | AUC |
|---|---|---|---|---|---|---|
CN vs. AD | ADNI1 | Early Fusion | \(0.950 \pm 0.016\) | \(0.952 \pm 0.023\) | \(\mathbf {0.948} \pm \mathbf {0.025}\) | \(0.947 \pm 0.013\) |
ADNI1 | Late Fusion | \(0.960 \pm 0.013\) | \(\mathbf {0.976} \pm \mathbf {0.030}\) | \(0.940 \pm 0.026\) | \(0.960 \pm 0.026\) | |
ADNI1 | Middle Fusion | \(\mathbf {0.962} \pm \mathbf {0.012}\) | \(\mathbf {0.976} \pm \mathbf {0.030}\) | \(0.948 \pm 0.026\) | \(\mathbf {0.969} \pm \mathbf {0.031}\) | |
ADNI2 & ADNI3 | Early Fusion | \(0.906 \pm 0.051\) | \(0.889 \pm 0.068\) | \(0.922 \pm 0.044\) | \(0.896 \pm 0.067\) | |
ADNI2 & ADNI3 | Late Fusion | \(0.885 \pm 0.063\) | \(0.813 \pm 0.120\) | \(0.957 \pm 0.037\) | \(0.906 \pm 0.064\) | |
ADNI2 & ADNI3 | Middle Fusion | \(\mathbf {0.963} \pm \mathbf {0.039}\) | \(\mathbf {0.907} \pm \mathbf {0.058}\) | \(\mathbf {0.965} \pm \mathbf {0.031}\) | \(\mathbf {0.949} \pm \mathbf {0.033}\) | |
CN vs. MCI | ADNI1 | Early Fusion | \(0.772 \pm 0.030\) | \(0.741 \pm 0.061\) | \(0.803 \pm 0.110\) | \(0.777 \pm 0.038\) |
ADNI1 | Late Fusion | \(0.770 \pm 0.040\) | \(\mathbf {0.697} \pm \mathbf {0.063}\) | \(0.843 \pm 0.096\) | \(0.763 \pm 0.035\) | |
ADNI1 | Middle Fusion | \(\mathbf {0.788} \pm \mathbf {0.024}\) | \(0.692 \pm 0.046\) | \(\mathbf {0.884} \pm \mathbf {0.036}\) | \(\mathbf {0.805} \pm \mathbf {0.041}\) | |
ADNI2 & ADNI3 | Early Fusion | \(0.739 \pm 0.054\) | \(0.676 \pm 0.101\) | \(0.802 \pm 0.102\) | \(0.766 \pm 0.052\) | |
ADNI2 & ADNI3 | Late Fusion | \(0.717 \pm 0.054\) | \(0.559 \pm 0.125\) | \(\mathbf {0.876} \pm \mathbf {0.071}\) | \(0.733 \pm 0.071\) | |
ADNI2 & ADNI3 | Middle Fusion | \(\mathbf {0.753} \pm \mathbf {0.036}\) | \(\mathbf {0.716} \pm \mathbf {0.047}\) | \(0.790 \pm 0.046\) | \(\mathbf {0.793} \pm \mathbf {0.037}\) |
