Table 1 Measured segmentation time per patient obtained by using CNN-based V-Net, ViT-based UNETR, and FS using CPU. GPU running time is shown in the ().
From: Comparative validation of AI and non-AI methods in MRI volumetry to diagnose Parkinsonian syndromes
CNN (s) | ViT (s) | FS (s) | |
|---|---|---|---|
Midbrain | \(8.5739\pm 2.50\;(0.5827\pm 0.17)\) | \(174.37\pm 10.81\;(7.5817\pm 0.47)\) | \(1698\pm 0.144\) |
Pons | \(8.5385\pm 2.35\;(0.5803\pm 0.16)\) | \(207.05\pm 46.46\;(9.2242\pm 2.02)\) | |
V3 | \(8.5341\pm 2.35\;(0.5800\pm 0.16)\) | \(178.25\pm 10.35\;(7.7525\pm 0.45)\) | \(14,037\pm 1.5\) |
Caudate | \(8.4590\pm 2.35\;(0.5749\pm 0.16)\) | \(176.20\pm 10.35\;(7.6610\pm 0.23)\) | |
Putamen | \(8.5561\pm 2.50\;(0.5815\pm 0.17)\) | \(179.63\pm 10.81\;(7.8112\pm 0.47)\) | |
Pallidum | \(8.6032\pm 2.65\;(0.5847\pm 0.18)\) | \(186.32\pm 21.39\;(8.1019\pm 0.93)\) | |
Total | \({51.26\pm 2.50\;(3.48\pm 0.17)}\) | \({1101.82 \pm 22.31\;(48.14 \pm 0.97)}\) | \({15,735\pm 1.07}\) |
