Histogram analysis- a useful tool for tissue characterization in brain CT

Document Type : Conference Proceedings

Authors

1 Medical imaging Research Centre, Shiraz University of Medical Sciences, Shiraz 7193635899, Iran

2 Ongil, 79 D3, Sivaya Nagar, Reddiyur Alagapuram, Salem 636004. India

3 Department of Instrumentation and Applied Physics, Indian Institute of Science, Bangalore 560012, India Professor

Abstract

Introduction: Pixel value in computed tomography (CT) gives the average linear attenuation coefficient of the scanned material in the path of the x-ray beam, being normalized to that of water. It is known that attenuation coefficient or HU value is a function of the chemical characteristic of the material and of the x-ray energy. The CT image shows the HU value by a shade of gray in each pixel. Human eye is often not able to discriminate gray shadows; whose grayness are close to each other. Therefore, when two different tissues have very similar chemical characteristics they may not be distinguishable in the CT image by human eyes. For this reason, in CT image, it is difficult to distinguish between brain tumor and ischemia, in the hypodense area. The aim of this study is to use histogram analysis, based on the frequency distribution of the HU values in brain CT images to evaluate the possibility of distinguishing tumor from ischemia.
Materials and Methods: We collected 150 axial slices of brain CT, from known hypodense areas belonging to ischemia and tumor with similar appearance. They contained 75 samples of each type. Minimum, maximum, and mean HU values of ROI inside the lesions were used to make the histogram, cumulative and probability percentage distribution. We used 10 unknown cases, 5 tumors and 5 ischemia, to evaluate our finding.
Results: Bar chart and linear diagram histogram show that the minimum, maximum and mean of HU values are higher in ischemia cases. Also, the median of the cumulative distributions of minimum, maximum and mean appear at higher HU values for ischemia. The probability of occurrence of ischemia cases increases with increasing HU values of minimum, maximum and mean while the behavior of the tumor cases is completely reverse. We have done the same data analysis for 10 unknown cases. Our predictions match in 80% of the cases, with the radiologist’s conclusion, who were unaware of our findings.
Conclusion: Histogram analysis may be used as an effective tissue characterization tool in non-enhanced brain CT images. This method could distinguish hypodense areas of ischemic lesion from tumor, as is seen, for the first time, in the present study.

Keywords