Department of Radiology

Image Processing and Analysis Laboratory

 

 

The Image Processing and Analysis Lab is an integrated resource in the Department of Radiology for education, clinical support, and research purposes. The lab currently has five advanced computer workstations with a variety of consistently updated commercial and public image processing and analysis software packages:

 

Matlab 7.0a

 

IDL 7.0

 

Analyze 7.0

 

GraphPad Prism 4

 

ImageJ 1.38x

 

The laboratory is capable of developing customized image analysis and visualization methods for quantitative measurements in small animal imaging research and clinical radiological research in the Department of Radiology as well as in other collaborative research using image analysis at UTHSCSA. The laboratory is open to the graduate program students, residents, research and clinical faculty of the department so that they can make use of the image processing software that is not available elsewhere on campus. Currently, Qi 'Chris' Peng, Ph.D., assistant professor in Radiology, is the director of the Image Processing and Analysis Lab. There are also two Ph.D. students and a senior research image analyst working in this lab. We welcome collaborative research projects.

 

 

Q-Fat

 

 

Q-Fat is automated abdominal adipose tissue quantification software for MRI images. Abdominal adipose tissue, particularly, visceral adipose tissue is a major indicator of the metabolic conditions linking to several diseases, including type 2 diabetes, coronary heart disease, hypertension, and stroke. Adipose tissue is most commonly quantified on MRI images using a manual contour tracing method, which is very time-consuming and inaccurate.


We have developed this automated software package to achieve automated and accurate fat quantification. Generally, four steps are taken for the processing of each image. First, intensity correction is performed, which iteratively performs polynomial field correction to find the minimal entropy of the image grayscale histogram. Second, an optimized fat segmentation algorithm based on fuzzy c-means clustering is used to exclude non-fat pixels from consideration. Third, a threshold value is automatically calculated to generate a fat-only image, based on which fat quantification can be performed. In such an image, all pixels with intensity value lower than the specified threshold value are set to zero (dark) on the fat-only image. In the fourth step, three contours are automatically generated: 1) the contour at the external abdomen; 2) the contour at the interior boundary of the subcutaneous fat; 3) the contour that includes all the visceral adipose tissue. An advanced contour generation/modification tool is also provided if more contours or modification on automatic contours are needed. Fat quantification is then performed based on the generated fat-only images and the contours. The resultant text file is saved and can be retrieved from the program interface, or later after fat quantification. Overall, it takes approximately two minutes to process ten abdominal MRI slices accurately.