The Project Detail
Title: Automatic diagnostic of mammographic images using wavelet transform and artifical neural networks
Student: Alireza Sheikh Hassani
Project Level: The Final Thesis in MS. level
Supervisor: Hamid Abrishami Moghaddam
Date: 3/1/2002
Abstract:
  Clustered microcalcification are the most important indicators of breast cancer in mammograms.they are calcified regions appearing as small bright objects in mammograms. because of low contrast characteristic of mammographic images,microcalcification detection and diagnostic ptocess are difficult and time- consuming tasks for radiologists. this is the reason why a CADsystem would be useful to help the radiologists to reduce errorous diagnostic risk. the goal of this project is to develop a microcalcification detection and diagnostic algorithm for digital mammograms. the designed method consists of three main steps. in the first step, we use wavelet transform and statistical measures to extract 4 features per pixel. a three layer feedforward neural network for detection of suspicious pixels forming a microcalcification.in the second step uses these features, we define 18 features and use them as input to second neural network to extract microcalcification objects. finally a third neural network, uses five features to classify microcalcification clusters in to benign and malignant classes. expermental results show acceptable performance of this algorithm.
Keywords: mammography;microcalcification;wavelet transforms;artifical neural networks;CAD

Click to download a pack to correct viewing Persian in IE5 and above:
 Download site 1: Borna Rayaneh        
Download site 2: K.N.Toosi Web site