Table 3.

Linear regressions of reconstruction error metrics vs. SNR. Linear correlation coefficient (r), slope, and p-value from the linear regressions of the angular error with respect to PSOCT and the RMSE in PDFs with respect to fully sampled data as a function of SNR. CS reconstructions used an acceleration factor of R=3 and high-SNR training data.

		Angular error			RMSE in PDFs
		r	Slope (° per unit SNR)	p-value	r	Slope (% per unit SNR)	p-value
FS-DSI		−0.45	−0.25	0.051	-	-	-
FS-DTI		−0.77	−1.21	0.00011	-	-	-
CS training	CS method
Sample 1A	PCA	−0.67	−0.36	0.0017	−0.99	−0.96	4.4 × 10−17
	PINV	−0.84	−0.43	7.0 × 10−6	−0.99	−0.79	9.7 × 10−17
Sample 1B	PCA	−0.65	−0.30	0.0028	−0.99	−0.81	3.2 × 10−16
	PINV	−0.73	−0.28	0.00038	−0.99	−0.70	1.8 × 10−16
Sample 2	PCA	−0.81	−0.46	2.6 × 10−5	−0.99	−0.80	1.1 × 10−17
	PINV	−0.81	−0.35	2.4 × 10−5	−0.99	−0.66	6.9 × 10−17