. Author manuscript; available in PMC: 2016 Apr 28.

Published in final edited form as: Conf Proc IEEE Eng Med Biol Soc. 2013;2013:5398–5401. doi: 10.1109/EMBC.2013.6610769

TABLE I.

Classification rate (%) of the proposed algorithm compared to the original sparse codes (SC). Both methods were based on optimized SVM with (C = 1400, γ = 0.01).

Algorithm	Prediction	CN	ncMCI	cMCI	AD
Algorithm	Diagnosis	CN	ncMCI	cMCI	AD
Original SC	CN	44.7	44.7	5.3	5.3
	ncMCI	27.5	49.0	13.7	9.8
	cMCI	11.8	38.2	26.5	23.5
	AD	14.3	16.7	4.7	64.3
Proposed LSCG $(\begin{matrix} α = 0.5 \\ r = 0.21 \end{matrix})$	CN	63.2	28.9	2.6	5.2
	ncMCI	27.5	52.9	7.8	11.8
	cMCI	8.8	35.3	32.4	23.5
	AD	4.8	14.3	4.7	76.2