Table 3.

ALSFRS slope performance with most important features. RMSD and PCC are shown for all methods. Performance was obtained on the external test set (n = 731) using 10,000 bootstrap with resampling. FFNN was trained using the top 5 features, CNN and RNN were trained using ALSFRS questionnaire data and the top 4 features. The best values for each metrics are in bold.

Methods	RMSD	PCC
FFNN	0.534 (0.506–0.564)	0.414 (0.367–0.458)
CNN	0.530(0.501–0.559)	0.429(0.379–0.476)
RNN	0.544 (0.514–0.574)	0.375 (0.328–0.419)