Comparison of wavelet compressed MRI cineloop images versus originally acquired non-modified MRI images for detection of brain lesions

Date of Award


Document Type


Degree Name

Master of Medicine (MMed)


Imaging and Diagnostic Radiology (East Africa)


Brief background: Image compression has been widely accepted as the solution to the explosion of imaging data since the advent of multiaxial multisequence imaging. However, compromise on diagnostic quality secondary to this modification has been an issue. Many studies have been carried out looking at just noticeable differences but few have been undertaken to look at the difference in ability to detect pathology in original versus cineloop compressed images. This study was done to compare lesion detection ability of cineloop compressed images as compared to original images.

Objectives: A comparative study to evaluate detection of brain lesions in wavelet compressed Audio video Interleave (AVI) cineloop re-formatted MRI images as compared to originally acquired 16 bit digital MRI images.

Methods: This was a prospective comparative study. MRI brain images of consecutive non-selected patients referred for MRI brain examinations were compressed by means of a wavelet algorithm at ratios of 30:1(70%) and then converted into Microsoft AVI cine loop format using Image J freeware( Appendix 3).

All the sequences and all the sections of a particular examination (original and reformatted images) were reviewed on the same high resolution computer monitor by two independent reviewers (Appendix 2).

Analysis: Each of the observers captured their findings in a pre-defined form then comparison of lesion detection was done taking the original images as a gold standard. Sensitivity were calculated for evaluation of number of lesions, location of lesions, signal characteristics and image quality. As compared to the original brain MRI images the sensitivities in the cineloop compressed images for presence of lesions was 100%, Number of lesions present was 100%, location of largest lesion was 97%, T1 signal was 96%, T2 signal was 96%, GRE signal was 95%, FLAIR signal was 93%, Presence of restricted diffusion was 98%, post contrast enhancement was 98%, and edema surrounding lesion was 100%. There was reduction of subjective image quality with 79% of the compressed cineloop images scoring 1 less point than the original images.

Outcomes: This study proves cineloop compressed images at a a ratio of 1:30 do not reduce ability to detect lesions in terms of size, number, locations and signals. Despite reduced quality of images, the diagnosis remains largely unchanged. Hence, image compression can be applied to MRI Brain imaging for interpretation allowing ease of teleradiology, reduced storage costs and increased transmission speeds.

This document is available in the relevant AKU library