Table 4:

Average R² for predicted-versus-observed cell density for cross-validation^{^a}

	All Variables (23 Inputs)	Variables Selected by RF Importance (4 Inputs)	All Conventional Variables (6 Inputs)	Variables Selected by RF Importance: Conventional Only (4 Inputs)
Random forest	0.572	0.586	0.513	0.523
Linear	0.542	0.572	0.444	0.475
Neural network	0.265	0.460	0.382	0.379
Decision tree	0.301	0.325	0.376	0.376

↵a The columns list variables used to train the predictive model. “All Variables” is simply using all 23 imaging parameters of all 6 conventional sequences, whereas “RF Importance” and “RF Importance, Conventional” use the final 4 variable sets shown in Online Tables 4 and 5. A larger average R² indicated better performance.