Breast Cancer Diagnoser (BCD)

Detection and Classification of Microcalcifications in Digital Mammograms

Breast Cancer Diagnoser (BCD) v2.1

Developed by
John Zakos and Brijesh Verma, at CIRL, Griffith University, Gold Coast
Download a copy of BCD v2.1 now!

* Runs on Win 95 and NT.
* Ideally requires 16 megabytes ram, 10 megabytes free hard disk space.
* Once 'system.zip' has been downloaded, unzip 'system.zip' into a new directory and run 'bcd.exe' for system.

BCD system diagnoses microcalcifications in digital mammograms as benign or malignant. Additional features allow you to enlarge a selected area of the mammogram, zoom a selcted area of the mammogram, measure distances of areas in the mammogram and detect abnormalities. It is a very useful tool for diagnosing breast cancer and should be used as a doctor's second-opinion in the evaluation of a patient's condition.

In the development of the system, many features were extracted and tested in different combinations in an attempt to identify the best features that describe a microcalcification. The system extracts average histogram, entropy and standard deviation features from microcalcification areas. The backpropagation neural network uses these features to classify the microcalcification area as benign or malignant. Entropy and standard deviation features were modified to give a better classification rate than traditional entropy and standard deviation features.

Currently, the system has an 88.9% true positive classification rate.

BCD Main Screen