Table 1.

Effects of Gaussian noise on Analysis performances: Results from LCModel & ICA analyses of data generated with narrow LCModel/GAVA bases, with and without noise, shown. Ground-truth correlations of both LCModel and ICA estimates are high, and show little performance degradation due to noise. LCModel does not find Asp resonance in GAVA generated data, due to modeling differences, though ICA had no issue in resolving the metabolite's model resonance. (A) Results from analyses of data generated with LCModel bases; (B) Results from analyses of data generated with GAVA bases

	Asp	Cr	GABA	Glc	Gln	Glu	m-Ins	NAA	NAAG	PCh	s-Ins	Tau
(A)
Ideal data, LCM analysis	1.000	1.000	0.998	0.999	1.000	0.999	1.000	1.000	1.000	1.000	1.000	0.999
Noisy data, LCM analysis	0.983	0.997	0.969	0.986	0.993	0.985	0.994	0.998	0.995	0.997	0.995	0.979
Ideal data, ICA analysis	0.997	1.000	0.995	0.991	0.996	0.985	0.999	1.000	0.998	0.998	0.990	0.917
Noisy data, ICA analysis	0.986	0.999	0.968	0.981	0.991	0.981	0.997	0.999	0.997	0.997	0.984	0.960

	Asp	Cr	GABA	Gly	Gln	Glu	m-Ins	NAA	NAAG	PCh	s-Ins	Tau
(B)
Ideal data, LCM analysis	0	0.988	0.932	–	0.990	0.975	0.806	0.990	0.790	0.978	0.935	0.982
Noisy data, LCM analysis	0	0.987	0.906	–	0.986	0.970	0.811	0.988	0.783	0.975	0.929	0.978
Ideal data, ICA analysis	0.996	1.000	0.976	0.867	1.000	0.982	1.000	1.000	0.977	0.999	0.969	0.999
Noisy data, ICA analysis	0.985	0.999	0.754	0.869	0.992	0.989	0.996	0.999	0.969	0.998	0.968	0.995