Quantify relative cell disorganization from fluorescence microscopy image

Project Proposal ECE 533		Carmalyn Lubawy
					Melissa Skala

Problem
In patients at risk for oral cancer, the entire oral cavity is at risk 
for disease, thus it difficult to find the optimal site for biopsy.  
One way to identify biopsy sites is fluorescence microscopy.  
This technique allows for an evaluation of the morphological features of 
cells in the tissue.  Cancerous cells are more disorganized than normal 
cells.  We would like to quantify relative cell disorganization, so that 
a trained pathologist does not have to evaluate every image.  Our goal in 
this project is to develop an algorithm that gives relative differences 
in cell organization.

Proposed Approach
Four pairs of normal and cancerous images will be compared using this 
approach.  Our first approach is described below; we will make iterations 
on this approach to improve differences between normal and cancerous 
images.  

The first steps of our proposed algorithm will be to preprocess our 
images to remove graininess (median filter) and then to scale images 
appropriately using histogram equalization.  Finally, we will enhance 
structural features using an unsharp mask and thresholding functions.  

After the pre-processing steps, the relative disorganization of the 
images will be determined with Fourier transform analysis (Fitzke, 
Masters).  The full width at half max of line plots through the 
Fourier-transformed image will be used to differentiate normal and 
diseased tissues.  

References
Fitzke, F.W. et al.  Fourier Transform Analysis of Human Corneal 
Endothelial Specular Photomicrographs.  Exp. Eye Res. 65 1997.

Masters, B.R.  Diagnostic digital image processing of human corneal 
endothelial cell patterns.  SPIE 1360 1990

Christens-Barry W.A. et al.  Quantitative Grading of Tissue and Nuclei 
in Posture Cancer for Prognosis Prediction.  John Hopkins APL Technical 
Digest 18 1997.