Table 1.

Top part: A and B coefficients obtained for each metabolite when fitting the function A ​+ ​B ​× ​cos(2ϕ). The B coefficient is higher for neuronal than glial marker. The cost function was calculated for each subject using $c=\sqrt{\sum _{\text{ϕ}}(x_{\varphi }-y_{\varphi })²}$, $x_{\varphi }$ and $y_{\varphi }$ being the fitted and experimental data respectively. Here, the mean and standard deviation for c clearly increase in the case of MM, underlining the poor quality of the fit due to lower signal modulation as well as more pronounced inter-individual variations. Error corresponds to the standard deviation calculated over four animals. Bottom part: D_free and d_fiber values corresponding to the best fit for each metabolite to a basis of simulated signals using the infinite cylinder model. Error on D_free and d_fiber was estimated using Monte Carlo simulations. D_free appears underestimated whereas d_fiber is overestimated.

	METABOLITES
	tNAA (neuronal marker)	tCr	Tau	Ins (glial marker)	tCho (glial marker)	Lac	MM
Fitting to A ​+ ​B ​× ​cos(2ϕ)
Offset (A)	0.293 ​± ​0.004	0.235 ​± ​0.012	0.197 ​± ​0.011	0.228 ​± ​0.016	0.286 ​± ​0.018	0.130 ​± ​0.004	0.742 ​± ​0.030
Amplitude Modulation (B)	0.040 ​± ​0.004	0.036 ​± ​0.004	0.035 ​± ​0.005	0.030 ​± ​0.003	0.029 ​± ​0.003	0.028 ​± ​0.006	0.020 ​± ​0.012
B/A ratio	0.137 ​± ​0.014	0.155 ​± ​0.020	0.177 ​± ​0.024	0.132 ​± ​0.015	0.101 ​± ​0.014	0.212 ​± ​0.051	0.026 ​± ​0.016
Cost function	0.018 ​± ​0.006	0.027 ​± ​0.004	0.020 ​± ​0.004	0.024 ​± ​0.006	0.017 ​± ​0.006	0.029 ​± ​0.002	0.055 ​± ​0.019
Fitting to the infinite cylinders model
Dfree (μm2/ms)	0.20 ​± ​0.02	0.22 ​± ​0.04	0.25 ​± ​0.03	0.21 ​± ​0.03	0.17 ​± ​0.02	0.34 ​± ​0.06	/
dfiber (μm)	3.22 ​± ​0.29	3.89 ​± ​0.43	4.20 ​± ​0.26	4.14 ​± ​0.0.15	3.76 ​± ​0.30	4.80 ​± ​0.27	/