M4. Individual weights from a linear SVM trained on features including
demographic and clinical covariates, grey matter volume, lesion count and total
lesion load; based on T1-Gd MRI data of IFN-b treated patients (converters: 50,
non-converters: 49). Positive (negative) weights indicate that a larger feature
value will drive prediction towards conversion (non-conversion). For sex, being
female (male) corresponds to the positive (negative) axis. Weights are scaled
relatively to the largest individual weight, which is set equal to one. Note
that weights should only be interpreted qualitatively, not quantitatively.