Table 3.

Comparison between different models

		qMRI + clinical predictors		Clinical predictors only
Description	Model	Full Model C-index (CI)	Adjusted C-index (CI)	Full model C-index (CI)	Adjusted C-index (CI)
Seizure free (yes versus no) right-sided model	1	0.829 (0.77–0.89)	0.703 (0.69–0.72)	0.672 (0.59–0.75) (logistic regression)	0.582 (0.57–0.59)
	2	0.788 (0.72–0.86)	0.659 (0.65–0.67)
	3	0.701 (0.62–0.78)	0.675 (0.66–0.69)	0.598 (0.52–0.68) (random forest regression)	0.586 (0.58–0.60)
	4	0.710 (0.63–0.79)	0.689 (0.68–0.70)
Seizure free (yes versus no) left-sided model	1	0.762 (0.70–0.82)	0.664 (065–0.68)	0.680 (0.61–0.75) (logistic regression)	0.605 (0.59–0.62)
	2	0.740 (0.68–0.80)	0.642 (0.63–0.65)
	3	0.660 (0.59–0.73)	0.647 (0.64–0.66)	0.574 (0.50–0.65) (random forest regression)	0.571 (0.56–0.58)
	4	0.616 (0.54–0.69)	0.614 (0.60–0.61)
Engel I versus II–IV right-sided model	1	0.784 (0.71–0.86)	0.666 (0.65–0.68)	0.701 (0.62–0.78) (logistic regression)	0.623 (0.61–0.64)
	2	0.723 (0.64–0.81)	0.630 (0.61–0.65)
	3	0.594 (0.50–0.69)	0.588 (0.57–0.60)	0.620 (0.54–0.71) (random forest regression)	0.594 (0.58–0.61)
	4	0.619 (0.53–0.71)	0.611 (0.60–0.62)
Engel I versus II–IV left-sided model	1	0.821 (0.77–0.88)	0.729 (0.72–0.74)	0.733 (0.67–0.80) (logistic regression)	0.676 (0.67–0.69)
	2	0.813 (0.76–0.87)	0.722 (0.71–0.73)
	3	0.701 (0.63–0.78)	0.686 (0.68–0.70)	0.654 (0.58–0.73) (random forest regression)	0.651 (0.64–0.66)
	4	0.727 (0.66–0.80)	0.715 (0.71–0.72)

Adjusted C-index: concordance index adjusted using a random split method.

Models

1. Backward selection method, AIC as selection criteria—Logistic regression using selected variables.

2. Random Forest selection method—Logistic regression using selected variables.

3. Backward selection method—Random Forest regression using selected variables.

4. Random Forest selection method—Random Forest regression using selected variables.

CI = 95% confidence intervals.