Signal Power Estimation and its Novel Applications in Radiology as two birds with one stone

Background/Purpose

In radiology, MTF and SNR (or CNR) are known as established methods for evaluating the quality of phantom images. However, due to their complexity caused by multiple procedures involved in measuring of multiple ROI statistics such as mean, variance, etc., this study proposes a simplified signal power estimation method calculated directly from a pair of observed images. Then, in this research, MTF* and SNR* utilizing this proposed method are introduced as novel applications.

Methods

First, two observed phantom images are prepared by using the same imaging condition. Then, a signal power of the image is estimated from a covariance of a pair of the observed images using the following equation. V(x) = Cov(y₁, y₂) where V(·) and Cov(·) denote a variance and a covariance of · respectively. Here, x is a signal image vector, and y₁ and y₂ are observed image vectors obtained under the same imaging conditions, where y is constructed from the signal vector x summated the additive noise vector n, namely y = x + n. Finally, MTF* and SNR* are calculated using the estimated signal power V(x). The proposed methods are then compared with the conventional methods.

Results

While a direct comparison between conventional SNR and SNR* proved difficult due to differences in definition, both methods showed similar trends. On the other hand, the conventional MTF by Droege's method and the MTF* were almost in agreement with two significant figures as expected.

Conclusion

This investigation proposed simplified method for estimating signal power derived from the covariance of a pair of observed images. And this proposed method produced novel applications termed MTF* and SNR*, and has an more practical than the conventional method in terms of simplified procedure as two birds with one stone.