Shobhit Sharma MBBS, Presenter: Nothing to Disclose

Jayaram K. Udupa PhD, Abstract Co-Author: Nothing to Disclose

Yubing Tong PhD, Abstract Co-Author: Nothing to Disclose

Drew Avedis Torigian MD, MA, Abstract Co-Author: Nothing to Disclose

James Eric Schmitt MD, PhD, Abstract Co-Author: Nothing to Disclose

MR images present significant intensity variation across patients and scanners irrespective of imaging sequence used. The presence of thermal noise, image inhomogeneity and non-standardness of the MR image intensity gray scale poses challenges to MR image analysis. The purpose of this study is to demonstrate how a pattern classification strategy may be effective in liver MRI once the images are first “cleaned” as optimally as possible by correcting for intensity inhomogeneity and by accounting for non- standardness, so that image intensities are made to have consistent tissue specific meanings.

20 normal subjects and 30 cirrhosis patients who underwent T1-weighted (T1W) and T2-weighted (T2W) MRI of the abdomen were retrospectively studied. Image correction was performed for radiofrequency field inhomogeneities, and the live wire method was used to segment the liver and generate a liver mask. The liver mask was overlaid upon the corrected image and the resultant image was standardized for intensity. The standardized MRI intensity statistics were calculated within the liver mask.

The cumulative histogram for each imaging sequence for each group appeared chaotic before standardization but well-behaved after standardization. Mean intensity, standard deviation, and %coefficient of variation from these distributions without vs. with standardization, respectively, were as follows. Normal: In-phase T1W (301, 219, 73% vs. 2707, 479, 18%); Opposed-phase T1W (270, 190, 70% vs. 2665, 700, 26%); Fat-suppressed opposed-phase 3D T1W (297, 266, 90% vs. 2798, 475, 17%); T2W (166, 116, 70% vs. 979, 560, 57%); T2W-FS (165, 137, 83% vs. 1081, 512, 47%); Cirrhosis: In-phase T1W (196, 100, 51% vs. 2667, 553, 21%); Opposed-phase T1W (183, 100, 55% vs. 2773, 546, 20%); Fat-suppressed opposed-phase 3D T1W (165, 110, 67% vs. 2734, 568, 21%); T2W (215, 170, 79% vs. 1019, 570, 56%); T2W-FS (163, 161, 99% vs. 1064, 552, 52%).  

Intensity inhomogeneity correction and intensity standardization may be useful to make MR image intensities have a tissue-specific numeric meaning. This novel approach may be useful to improve the detection and characterization of cirrhosis on MRI.

MR intensity standardization may help to detect and quantify differences between normal and cirrhotic livers compared to non-standardized MRI.

Sharma, S, Udupa, J, Tong, Y, Torigian, D, Schmitt, J, Effect of MRI Intensity Standardization on Liver Tissue Characterization: A Comparison between Normal and Cirrhotic Livers.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14006852.html