Colorization of CT images to improve tissue contrast for tumor segmentation

  • Authors:

  • Marisol Martinez-Escobar;Jung Leng Foo;Eliot Winer

  • Affiliations:

  • 1620 Howe Hall, Virtual Reality Applications Center, Iowa State University, Ames, IA 50011, United States;1620 Howe Hall, Virtual Reality Applications Center, Iowa State University, Ames, IA 50011, United States;Mechanical Engineering and Human-Computer Interaction Department, 1620 Howe Hall, Virtual Reality Applications Center, Iowa State University, Ames, IA 50011, United States

  • Venue:
  • Computers in Biology and Medicine
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Segmenting tumors from grayscale medical image data can be difficult due to the close intensity values between tumor and healthy tissue. This paper presents a study that demonstrates how colorizing CT images prior to segmentation can address this problem. Colorizing the data a priori accentuates the tissue density differences between tumor and healthy tissue, thereby allowing for easier identification of the tumor tissue(s). The method presented allows pixels representing tumor and healthy tissues to be colorized distinctly in an accurate and efficient manner. The associated segmentation process is then tailored to utilize this color data. It is shown that colorization significantly decreases segmentation time and allows the method to be performed on commodity hardware. To show the effectiveness of the method, a basic segmentation method, thresholding, was implemented with and without colorization. To evaluate the method, False Positives (FP) and False Negatives (FN) were calculated from 10 datasets (476 slices) with tumors of varying size and tissue composition. The colorization method demonstrated statistically significant differences for lower FP in nine out of 10 cases and lower FN in five out of 10 datasets.