Application of despeckle filter local statistic mean variance (DsFlsmv) techniques towards image quality and anatomicalinformation on pancreatic ultrasound images

Abstract

Pancreatic ultrasound examination has a weakness, namely the presence of granular noise and disturbing image quality known as speckle noise. To be able to reduce speckle noise and improve the quality of pancreatic ultrasound images, this study used despeckle filter local statistic mean variance (DsFlsmv) filter technique.

This type of quasi-experimental study used a pretest posttest without control group design. The study sample was 32 volunteers. Qualitative analysis on image information begins with a visual assessment by 3 (three) radiologists then analyzed used Friedman test while quantitative analysis of image quality used Paired T-test.

Qualitative study of image quality shows differences in pancreatic ultrasound image information between pre-denoising, post-denoising size window 3x3 and post-denoising size window 5x5 images with a p-value < 0.001. The highest mean rank was obtained in the post-denoising size 3x3 images of 2.82. Quantitative study has different image quality (MSE and PSNR values) on pancreatic ultrasound post-denoising size 3x3 and post-denoising size window 5x5 images with p-value < 0.001. The lowest mean MSE value obtained in post-denoising size window 3x3 images is 5.063197 and the highest mean PSNR value in the post-denoising image size 3x3 images is 41.099753 dB.

The use of DsFlsmv technique with size window 3x3 can improve image quality and result the most optimal anatomical information on pancreatic ultrasound images.