Classification and Interpretability of Mild Cognitive Impairment Based on Resting-State Functional Magnetic Resonance and Ensemble Learning

Table 1

Partial ML-based studies with rs-fMRI.

Year	ML method	Subjects	Performance
2022 [20]	CNN	NC: 167, eMCI: 102, lMCI: 129, AD: 114	Average accuracy 89%
2022 [21]	AdaBoost	eMCI: 34, lMCI: 32	Accuracy 70%
2022 [22]	SVM	NC: 20, AD: 27	Accuracy (fMRI) 78.72%
			Accuracy (sMRI + rs-fMRI) 91.49%
2022 [23]	SVM	NC: 41, aMCI: 30, AD: 36	Accuracy (NC vs. aMCI) 68%
			Accuracy (NC vs. AD) 71%
2022 [24]	LDA	NC: 30, AD: 28	Accuracy 76.7%
2021 [25]	SVM	MCI-C: 14, MCI-NC: 41	Accuracy (fMRI) 83.5%
			Accuracy (sMRI + rs-fMRI) 83.5%
2021 [26]	SVM	MCI-C: 30, MCI-NC: 55, AD: 19	Accuracy (MCI-C vs. MCI-NC) 84.71%
			Accuracy (MCI-C vs. AD) 89.80%
2020 [27]	SVM	NC: 51, MCI: 66	Accuracy 85.5%
2020 [28]	SVM	NC: 20, SCD: 22	Accuracy 83.3%
2020 [29]	RF	NC: 83, MCI: 82	Accuracy 91.4%
2020 [30]	SVM	NC: 136, SMC: 46, eMCI: 83, MCI: 37, lMCI: 46, AD: 35	AUC (AD vs. NC) 0.87
2020 [31]	ANFIS	AD: 33, VD: 27, MXD: 15	Average accuracy 77.33%
2020 [32]	SVM	eMCI: 77, lMCI: 64	Accuracy 87.94%
2020 [33]	SVM	NC: 60, MCI: 39	AUC 0.9728
2019 [34]	ResNet-18	NC: 25, SMC: 25, eMCI: 25, lMCI: 25, MCI: 13, AD: 25	Average accuracy 97.88%
2019 [11]	SVM	NC: 49, MCI-NC: 69, MCI-C: 25, AD: 34	Accuracy (AD vs. MCI-C vs. MCI-NC) 67.6%
			Accuracy (NC vs. MCI-C vs. MCI-NC) 66%
			Accuracy (AD vs. NC vs. MCI-C vs. MCI-NC) 56.1%
2019 [35]	SVM	NC: 45, AD: 45	Accuracy 81.11%
2019 [36]	CNN	NC: 172, eMCI: 179	Accuracy 73.85%
2020 [37]	CNN	NC: 198, AD: 133	Accuracy 85.27%
2019 [38]	SVM	NC: 45, SCD: 39, aMCI: 45, AD: 38	Accuracy (AD vs. NC) 98.58%
			Accuracy (aMCI vs. NC) 97.76%
			Accuracy (SCD vs. NC) 80.24%
2019 [39]	SVM	NC: 24, eMCI: 24, lMCI: 24, AD: 24	Accuracy (eMCI vs. NC) 93.8%
			Accuracy (lMCI vs. NC) 95.8%
			Accuracy (AD vs. NC) 95.8%
			Accuracy (eMCI vs. lMCI) 87.5%
			Accuracy (lMCI vs. AD) 91.7%
2019 [40]	ELM	NC: 31 + 152, MCI: 31 + 132, AD: 33 + 81	In ANDI-2 cohort: Accuracy (AD vs. NC) 94.07%
			Accuracy (MCI vs. NC) 87.54%
			In the in-house cohort: Accuracy (AD vs. NC) 95.5%
			Accuracy (MCI vs. NC) 86.52%
2018 [41]	DAG network	NC: 34, AD: 34	Accuracy 95.59%
2018 [42]	SVM	MCI-C: 18, MCI-NC: 62	Accuracy 97%
2019 [43]	AE	NC: 79, MCI: 91	Accuracy 86.47%
2018 [44]	SVM	NC: 35, AD: 25	Accuracy 94.44%