Skip to main content

Table 1 Overview of diagnostic performances of single modalities (bSVM) and concatenated modalities (MKL) for the SN

From: A machine learning-based classification approach on Parkinson’s disease diffusion tensor imaging datasets

bSVM MKL
  BA [%] ROC-AUC Sens [%] Spec [%]   BA [%] ROC-AUC Sens [%] Spec [%]
FA 47.8 .42 47 48 FA + MD + AD + RD 49.4 41 44 60
MD 50.0 .54 55 42 06LDHs + 18LDHs + 26LDHs 56.1 60 54 56
AD 50.0 .44 40 47 06LDHk + 18LDHk + 26LDHk 58.1 52 56 41
RD 50.0 .54 48 41      
06LDHs 49.4 .54 40 44      
18LDHs 56.6 .57 49 41      
26LDHs 53.1 .53 51 60      
06LDHk 55.2 .52 51 56      
18LDHk 53.1 .53 60 63      
  1. Besides BA and ROC-AUC. Sens and Spec are listed to enhance the transparency of reported ROC-AUC results
  2. AD Axial diffusivity, BA Balanced accuracy, bSVM Binary Support vector machine, FA Fractional anisotropy, LDH Local diffusion homogeneity, MD Mean diffusivity, MKL Multiple-kernel learning, RD Radial diffusivity, ROC-AUC Receiver operator characteristics area under the curve, Sens Sensitivity, Spec Specificity, SN Substantia nigra