Table 9 The performance of the explainers on different modalities (First Layer).

From: A multilayer multimodal detection and prediction model based on explainable artificial intelligence for Alzheimer’s disease

Explainer	No. of features	Model	Exp. (%)	Precision	Recall	Accuracy	F1-score	Exp. measures
Cognitive scores based	12	DT	96 (91.43)	0.913	0.912	0.912	0.912	NL = 29; S = 57
Cognitive scores based	12	FURIA	97 (92.38)	0.910	0.909	0.909	0.909	NR = 17
Genetics based	5	DT	61 (58.10)	0.553	0.555	0.555	0.553	NL = 35; S = 69
Genetics based	5	FURIA	63 (60.00)	0.556	0.559	0.559	0.553	NR = 8
Lab tests based	41	DT	39 (37.14)	0.408	0.407	0.407	0.408	NL = 155; S = 309
Lab tests based	41	FURIA	53 (50.48)	0.469	0.494	0.494	0.460	NR = 6
Medical history based	27	DT	83 (79.05)	0.495	0.496	0.496	0.494	NL = 173; S = 331
Medical history based	27	FURIA	67 (63.81)	0.521	0.519	0.519	0.517	NR = 10
MRI based	8	DT	64 (60.95)	0.472	0.471	0.471	0.468	NL = 58; S = 115
MRI based	8	FURIA	51 (48.57)	0.525	0.525	0.525	0.524	NR = 5
Neurological exams based	12	DT	47 (44.76)	0.387	0.451	0.451	0.335	NL = 15; S = 29
Neurological exams based	12	FURIA	47 (44.76)	0.413	0.467	0.467	0.350	NR = 3
Neuropsychological battery based	35	DT	87 (82.86)	0.733	0.732	0.732	0.732	NL = 85; S = 169
Neuropsychological battery based	35	FURIA	85 (80.95)	0.771	0.771	0.771	0.770	NR = 22
PET based	3	DT	60 (57.14)	0.614	0.612	0.612	0.610	NL = 22; S = 43
PET based	3	FURIA	66 (62.86)	0.621	0.618	0.618	0.616	NR = 6
Physical exams based	10	DT	48 (45.71)	0.291	0.443	0.443	0.304	NL = 13; S = 25
Physical exams based	10	FURIA	49 (46.67)	0.257	0.461	0.461	0.297	NR = 5
Symptoms based	27	DT	49 (46.67)	0.425	0.449	0.449	0.383	NL = 32; S = 63
Symptoms based	27	FURIA	49 (46.67)	0.430	0.446	0.446	0.389	NR = 3
Vital signs based	8	DT	36 (34.29)	0.369	0.417	0.417	0.367	NL = 79; S = 157
Vital signs based	8	FURIA	45 (42.86)	0.360	0.453	0.453	0.329	NR = 2

Exp. measures, explainability measure; NL, number of leaves; S, size of the tree; NR, number of rules; Exp. (%), the number of explained cases (percentage of coverage); values in bold indicate the best performance.

Back to article page

Table 9 The performance of the explainers on different modalities (First Layer).

Search

Quick links