Breast cancer is a leading cause of cancer deaths among women. However the early detection for cancerous tissues can help reduce the mortality rates. Until now, the best way for detecting the cancerous tissues is via the use of mammographic images. The problem with mammography is the large number of screenings that produces a large number of mammography images, those images need to be thoroughly examined by radiologists and doctors which is both financially consuming and time inefficient. Added to that false detection can cause a high number of unnecessary biopsies. To overcome the previously mentioned problems a lot of research has been dedicated to developing computer aided detection applications that can help reduce the work load and improve the detection rates. This project is an example of the ongoing research regarding that topic and is conducted in the image and video systems lab of the Korean advanced institute of technology.
2. Project Goals
- Single-view mass detection with an improved classification design.
- Improving mass detection using two views (CC/MLO).
- Micro-calcification detection from a single-view.
- Integrated mass and micro-calcification detection.
3. Hardware Configuration
Figure 1 show the demo KAIST-Mammo CAD system. As the figure shows the demo setup uses two monitors with QHD (2550X1440) as well as a trackball.
4. Software Architecture and Functionality
4.1. Software Architecture:
Figure 2 shows the basic class diagram showing the basic skeleton for the KAIST-Mammo CAD system. The system was developed and built using Microsoft Visual C++ 10.0 based on MFC.
4.2. KAIST-Mammo CAD System Functionality:
Figure 3 show the main pane for the demo KAIST-Mammo CAD System, by adjusting the controllers in this pane we can perform the following functionalities:
- Single-View Mass Detection.
- Two-View Mass Detection.
- Micro-calcification Detection.
- Combined Mass and Micro-calcifications Detection.
- Viewer Aiding Tools.
Those functionalities will be further discussed in the following subsections.
4.2.1. Single-View Mass Detection
Figure 4 shows the configuration for the single-view mass detection. The demo KAIST-Mammo CAD System provides the functionality to detect a mass from one of the views (CC or MLO) in mammograms. The system detects mass areas by marking a blue line around the detected region. The detected mass regions can be adjusted by adjusting the threshold scroll bar (confidence values).
4.2.2. Two-View Mass Detection
Figure 5 shows the configuration for the two-view mass detection. The demo KAIST-Mammo CAD System provides the functionality to detect a mass from both of the mammography views (CC and MLO) for the same breast laterality. The system detects pairs detected ROIs one from each view and classify it as a mass accordingly.
4.2.3. Microcalcification Detection
Figure 6 shows the configuration for the micro-calcification detection. The demo KAIST-Mammo CAD System provides the functionality to detect micro-calcification from one mammography view (CC or MLO). The system detects micro-calcification cluster regions by marks them with blue triangles.
4.2.4. Combined Mass and Micro-calcification Detection
Figure 7 shows the configuration for the combined mass and micro-calcification detection. The demo KAIST-Mammo CAD System provides the functionality to detect both masses and micro-calcification clusters from a single mammography view (CC or MLO). In this configuration the system gives higher weights for mass regions with micro-calcifications inside them.
4.2.5. Viewer Aiding Tools
The following tools were also developed in order to help the radiologist while viewing the mammograms.
- Zoom in/out, magnification and moving.
- Intensity change and inversion.
- Control the size of ROIs(mass, calcification)
- Contact Person: Prof. Yong Man Ro (firstname.lastname@example.org)