public class AlgorithmTwoMRIImagesSNR
extends AlgorithmBase

This module implements the National Electrical Manufacturers Association (NEMA) standard for SNR estimation in diagnostic MRI. This method requires a perfect geometrical registration of 2 MRI images scanned sequentailly with less than five minutes elapsed time from the end of the first scan to the beginning of the second scan. A simple 2 step process is used: 1.) Determine the standard deviation of the difference image over the same VOI region used for the signal mean difference variance = (1/(VOI pixel number - 1)) * sum over VOI region of (image[i] - image2[i])**2 difference standard deviation = square root(difference variance)

2.) SNR for a signal VOI = square root(2) * (voi mean)/difference standard deviation The factor square root(2) is required because the standard deviation used is from the difference image rather than one of the original images The contrast to noise ratio is simply SNR for VOI 1 - SNR for VOI 2.

An optional registration may be performed before SNR. In this registration image2 is registered to image. AlgorithmRegOAR2D is used with the cost function being the only registration parameter the user can vary in the dialog box. Correlation ratio is the default cost function, but the user can also select least squares, normalized cross correlation, or normalized mutual information. The SNR will be performed with the registered image2 rather than on the original prebleached image.

Note that even for 2 perfectly aligned images the results differ substantially for processing without and with registration. In 1 example for 2 aligned images without registration the standard deviation equalled 12.3 and the SNR equalled 8.44. With registration the standard deviation equalled 7.66 and the SNR equalled 13.6.

The code here assumes the presence of 1 or 2 VOI regions - a required signal region and optional second signal region. The VOIs must all be placed in the the same image. Radio buttons in the dialog box are used to select a red signal 1 or a green signal 2 VOI. Either an ellipse VOI, rectangle VOI, or polyline VOI will be selected from the top MIPAV toolbar. There is no need to hit the NEW_VOI button.

Reference: 1.) PhD. Thesis Signal and Noise Estimation From Magnetic Resonance Images by Jan Sijbers, Universiteit Antwerpen, Department Natuurkunde, Section 7.1.2 The NEMA standard, p. 49.

2.) Acceptance Testing of Magnetic Resonance Imaging Systems, AAPM Report No. 34, Published by the American Assoication of Physicists in Medicine by the American Institute of Physics, March, 1992 (Reprinted from Medical Physics, Vol. 19, Issue 1, 1992).