Does Precision Imaging (PI) improve image quality in breast ultrasound?

Abstract

Objective To determine the effect of the PI speckle reduction algorithm on image quality in breast ultrasound. Introduction

Ultrasound is an integral tool used to identify breast cancer specifically in populations with mammographically dense breasts. One of the drawbacks of ultrasound is its tendency to produce a grainy image due to speckle noise. A granulated image is generally held to be of poorer quality, and this interference makes image contrast and spatial resolution difficult to appreciate. Not only does this reduce the quality of the image, but it also makes lesion characterization particularly difficult. Image enhancement methods that boost detection of low contrast objects, small high contrast targets and small differences in image brightness require investigation. Precision Imaging (PI) introduced by Toshiba is a post-processing speckle reduction algorithm. Toshiba claims that its software improves diagnostic confidence through enhanced structure definition and reduced noise (Azar, 2011; Toshiba, 2009) Material and methods

The images included in this research were patients aged from 20 to 85 years who attended The Sydney Breast Clinic from October 2010 to July 2011. The commercial ultrasound scanner used for image acquisition was the Toshiba AplioMX, Model SSA-780A, (Toshiba Medical Systems, Otawara-shi, Tochigi-ken, Japan) with compact linear transducers 15-7MHz (PLT-1204BT) and 12-5 MHz (PLT-805AT). The image considered to best represent the area of interest was recorded without PI (L0), then with all other 3 levels of PI, namely Precision 1 (L1), Precision 2 (L2) and Precision 3 (L3). 247 sets of de-annotated images (92 with normal breast parenchyma and 155 showing benign lesions) were selected. The images were evaluated by one radiologist and five sonographers dedicated to breast imaging. They scored each lesion from 1 and 4 with 1 being the most preferred image and 4 being the least preferred. Results

The Friedman test showed overall that the results were significant in favour of more speckle reduced images. That is, all except one observer preferred L2 PI and L3 PI. A Wilcoxon T paired analysis also demonstrated a statistically significant difference (p<0.001). Conclusion

Image quality can be improved with an increased PI level, and the majority of observers prefer the most speckle-reduced image over others.