Table 3. AUC analysis for WM detection using the myelin predictions.

Model	Loss	Lesion Mask	Cohort(s) used	AUC	CI Low	CI High
Linear GAM	RMSE	n	all	0.832	0.822	0.832
	MAE	n	all	0.833	0.823	0.834
Segmentation Regression	RMSE	n	all	0.854	0.844	0.863
	MAE	n	all	0.856	0.846	0.866
Markov- GAM	RMSE	n	all	0.826	0.815	0.826
	MAE	n	all	0.827	0.816	0.827
Markov GAM	RMSE	n	all	0.844	0.835	0.855
	MAE	n	all	0.843	0.833	0.853
Myelin Feature	-	n	all	0.883	0.874	0.891
Segmentation Regression+	MAE	y	day 7	0.904	0.888	0.918
Markov+ GAM	MAE	y	day 7	0.910	0.895	0.924
Myelin Featuree	-	y	day 7	0.919	0.905	0.932

Note: Bold highlights the best results. CI: confidence interval, 95%. Segmentation Regression+: segmentation regression plus lesion masks; Markov- GAM: GAM based on T2 signal intensity alone; Markov+ GAM: Markov GAM plus lesion masks.