Quality assessment of compressed cardiac MRI. Effect of lossy compression on computerized physiological parameters

With the increasing number of medical imaging modalities, compressing digital medical images can facilitate their transmission and storage. This paper evaluates and compares two lossy compression methods: JPEG and wavelet, applied to the end diastolic phase slices of a cardiac MRI acquisition. First, the qualitative image quality was evaluated with clinical criteria and was performed by radiologists. The compression image quality was then not judged satisfactory over 22:1 for JPEG and 32:1 for wavelets. Secondly, we evaluated the effect of these lossy methods on physiological parameters, such as left ventricular mass and end-diastolic volume, calculated by an automated segmentation algorithm. The wavelet coding allows a compression ratio of 67:1, while JPEG is limited to 30:1. The wavelet compression performance was superior to JPEG at high and low compression ratios. We conclude that limiting the compression ratios to those defined by the psychovisual assessment secures the validity of physiological parameters in cardiac MRI