Computer Aided Diagnosis Module V1. for NAVICAD Project
Contract Number: 3SEE/30.06.2014,
Project Code: EEA-J RP-RO-NO-2013-1-0123
Project Title:
This module is using third party Matlab functions:
“isocontour”, developed by Dirk-Jan Kroon (
“polygeom”, developed by H.J. Sommer, (

Confocal laser endomicroscopy (CLE) is becoming a popular method for optical biopsy of digestive mucosa for both diagnostic and therapeuticprocedures. Computer aided diagnosis of CLE images, using image processing and fractal analysis can be used to quantify the histological structures in the CLE generated images. The aim of this study is to develop an automatic diagnosis algorithm of colorectal cancer (CRC), based on fractal analysis and neural network modeling of the
CLE-generated colon mucosa images.

We retrospectively analyzed a series of 1035 artifact-free endomicroscopy images, obtained during CLE examinations from normal mucosa (356 images) and tumor regions (679 images). The images were processed using this computer aided diagnosis (CAD) medical imaging system in order to obtain an automatic diagnosis. The database with images is freely available for testing on this site.

The CAD application includes image reading and processing functions, a module for fractal analysis, grey-level co-occurrence matrix (GLCM) computation module, and a feature identification module based on the Marching Squares and linear interpolation methods.

A two-layer neural network from MATLAB toolbox used to automatically interpret the imaging data and diagnose the pathological samples based on the fractal dimension and the characteristic features of the biological tissues.

Normal colon mucosa is characterized by regular polyhedral crypt structures whereas malignant colon mucosa is characterized by irregular and interrupted crypts, which can be diagnosed by CAD. For this purpose, seven geometric parameters were defined for each image: fractal dimension, lacunarity, contrast correlation, energy, homogeneity, and feature number. Of the seven parameters only contrast and feature number were significantly different between normal and cancer samples.

Next, a two-layer feed forward neural network was used to train and automatically diagnose the malignant samples, based on the seven parameters tested. The neural network operations were cross-entropy with the results: training: 0.53, validation: 1.17, testing: 1.17, and percent
error, resulting: training: 16.14, validation: 17.42, testing: 15.48. The diagnosis accuracy error was 15.5%.

matlab code

Multiple volume image archive:

Set of 1035 jpeg file stacks of normal and cancer CLE images. To open, click and download each .rar folder. Unzip. rar folder and visualize individual jpeg image with an image processing software i.e. ImageJ

CLE images for test – normal – volume 1

CLE images for test – normal – volume 2

CLE images for test – normal – volume 3

CLE images for test – normal – volume 4

CLE images for test – cancer – volume 1

CLE images for test – cancer – volume 2

CLE images for test – cancer – volume 3

CLE images for test – cancer – volume 4

CLE images for test – cancer – volume 5

CLE images for test – cancer – volume 6

CLE images for test – cancer – volume 7

Results of the fractal analysis with the CAD module:

PLOS Stefanescu fractal analysis results

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>