Table 4.

Mean plus or minus standard deviation for eloquent class TPR, specificity, F1 and AUC for the JHH cohort, where the number of subjects who performed each task is shown in the first column. The final column shows the FDR corrected p-values for the associated t-scores where we compare AUC between our method against each baseline.

Task	Method	Eloquent TPR	Specificity	F1	AUC	t-score	p-value
Language	MTGNN	0.75 ± 0.011	0.72 ± 0.01	0.74 ± 0.013	0.76 ± 0.013
(N = 62)	FCNN	0.68 ± 0.014	0.63 ± 0.016	0.67 ± 0.013	0.70 ± 0.015	11.56	3.8 × 10−30
	RF	0.49 ± 0.034	0.65 ± 0.027	0.59 ± 0.029	0.61 ± 0.035	12.11	5.7 × 10−33
	SVM	0.46 ± 0.017	0.55 ± 0.019	0.45 ± 0.02	0.52 ± 0.012	50.76	≈ 0
Finger	MTGNN	0.85 ± 0.014	0.83 ± 0.016	0.82 ± 0.013	0.83 ± 0.015
(N = 38)	FCNN	0.77 ± 0.019	0.65 ± 0.016	0.73 ± 0.019	0.75 ± 0.017	8.36	2.7 × 10−16
	RF	0.48 ± 0.039	0.66 ± 0.028	0.57 ± 0.034	0.60 ± 0.029	24.22	1.7 × 10−128
	SVM	0.55 ± 0.02	0.54 ± 0.021	0.53 ± 0.015	0.54 ± 0.014	43.48	≈ 0
Foot	MTGNN	0.81 ± 0.023	0.81 ± 0.021	0.79 ± 0.019	0.78 ± 0.025
(N = 18)	FC-NN	0.71 ± 0.023	0.62 ± 0.025	0.68 ± 0.024	0.73 ± 0.025	9.32	5.5 × 10−20
	RF	0.45 ± 0.044	0.67 ± 0.038	0.51 ± 0.039	0.66 ± 0.047	10.58	2.0 × 10−25
	SVM	0.53 ± 0.028	0.57 ± 0.023	0.49 ± 0.025	0.54 ± 0.021	25.63	1.2 × 10−143
Tongue	MTGNN	0.82 ± 0.015	0.81 ± 0.012	0.82 ± 0.014	0.80 ± 0.014
(N = 41)	FC-NN	0.83 ± 0.019	0.80 ± 0.011	0.83 ± 0.018	0.80 ± 0.019	−0.91	0.82
	RF	0.38 ± 0.028	0.65 ± 0.029	0.52 ± 0.024	0.60 ± 0.031	18.96	3.5 × 10−79
	SVM	0.58 ± 0.021	0.51 ± 0.022	0.50 ± 0.025	0.53 ± 0.015	37.69	1.34 × 10− 309