Assisted breast tissue abnormality differentiation using magnetic resonance images
Generally, a complete set of breast Magnetic Resonance Imaging (MRI) images consist of at least 2000 pre- and post-processed images. That is very time consuming if the images are analyzed manually. By adopting the computer aided detection (CAD-e) and diagnosis (CAD-x) system in breast MR images, the...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2012
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Generally, a complete set of breast Magnetic Resonance Imaging (MRI) images consist of at least 2000 pre- and post-processed images. That is very time consuming if the images are analyzed manually. By adopting the computer aided detection (CAD-e) and diagnosis (CAD-x) system in breast MR images, the analyzing time can be effectively reduce and it can provide the priceless clinical decision supporting information, for instance, the tumour position and its malignancy. Numerous approaches have been proposed to detect and diagnose the MR breast images, for instance, Brown et al. (2000), Szabo et al. (2003), Nermin et al. (2005) and etcetera. However, they experience several drawbacks such as manually drawn breast region of interest (ROI) and machine dependence. A novel CAD-x auto probing system is developed to overcome the mentioned obstacles. The proposed method has been applied on digital image and communications in medicine (DICOM) breast MRI datasets images. The outcomes of the proposed scheme are independently verified by two radiologists. The designed standalone CAD-x system improves the elimination of noise, refines breast region of interest (ROI), and detects the breast lesion with minimal false positive detection. By using the wash-out technique, the detected breast lesions are then classified and colourised in three categories which are benign, suspicious or malignant according to the pharmacokinetic models describing the exchange of contrast agents (CA) molecules between tissue compartments over time. In order to enhance the visualization, the entire analyzed breast ROI is constructed into 3D view. The system output shows the proposed scheme is able to identify and classify the breast cancer with a satisfactory reliability. By comparing with other two existing methods developed by Kelcz et al. (2001) and Kneeshaw et al. (2006), it is capable to provide higher detection accuracy, sensitivity, and specificity. |
|---|