Table 4. Mean ± standard deviation (min to max) cross-validated error and 95% confidence interval (min to max, last row) of cellularity model accuracies.

Model	Loss	R2	Pearson Correlation	Residual Correlation
Linear GAM	RMSE	0.31 ± 0.23	0.66 ± 0.11	0.84 ± 0.08
		(−0.36 to 0.51)	(0.48 to 0.84)
	MAE	0.29 ± 0.30	0.66 ± 0.11	0.77 ± 0.09
		(−0.58 to 0.56)	(0.48 to 0.84)
Segmentation Regression	RMSE	0.44 ± 0.17	0.69 ± 0.12	0.73 ± 0.08
		(0.08 to 0.66)	(0.44 to 0.85)
	MAE	0.39 ± 0.24	0.69 ± 0.12	0.57 ± 0.09
		(−0.12 to 0.70)	(0.43 to 0.84)
Segmentation Regression+	RMSE	0.55 ± 0.05	0.79 ± 0.03	0.58 ± 0.09
		(0.50 to 0.62)	(0.75 to 0.83)
	MAE	0.52 ± 0.07	0.79 ± 0.03	0.46 ± 0.13
		(0.43 to 0.61)	(0.75 to 0.82)
Markov-GAM	RMSE	0.35 ± 0.22	0.68 ± 0.12	0.82 ± 0.08
		(−0.26 to 0.59)	(0.45 to 0.86)
	MAE	0.31 ± 0.29	0.67 ± 0.11	0.70 ± 0.10
		(−0.48 to 0.61)	(0.45 to 0.84)
Markov GAM	RMSE	0.36 ± 0.26	0.68 ± 0.12	0.76 ± 0.08
		(−0.41 to 0.61)	(0.38 to 0.84)
	MAE	0.36 ± 0.29	0.69 ± 0.12	0.70 ± 0.10
		(−0.47 to 0.66)	(0.42 to 0.85)
Markov GAM+	RMSE	0.54 ± 0.07	0.79 ± 0.06	0.57 ± 0.14
		(0.49 to 0.64)	(0.72 to 0.86)
	MAE	0.54 ± 0.07	0.80 ± 0.06	0.50 ± 0.18
		(0.49 to 0.66)	(0.73 to 0.87)
Cellularity Feature	-	-	0.60	-
			(CI: 0.59 to 0.62)

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