Digital infrared thermography analysis for breast cancer using image textures features / Norlailah Lanisa
The number of breast cancer has been increasing over the decade. Multiple screening tools have been used to detect the cancer, such as mammogram, ultrasound, MRI, and thermography. Mammogram is the gold standard method for breast cancer screening. However, thermography screening method has caught th...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2017
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/8027/7/norlailah.pdf http://studentsrepo.um.edu.my/8027/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The number of breast cancer has been increasing over the decade. Multiple screening tools have been used to detect the cancer, such as mammogram, ultrasound, MRI, and thermography. Mammogram is the gold standard method for breast cancer screening. However, thermography screening method has caught the attention to be used as alternative screening tools due to its non-contact procedure as well as it is convenient for patients below 40 years old. In this study, thermography image will be used as the input image. The proposed method start by overlaying the binary ground truth (GT) mask with the greyscale thermography image to eliminate the background image. Then followed by manual extraction of the left and right breast into two similar size individual region of interest (ROI). Later, the ROI images contrast were improved by using histogram equalization (HE). From the equalized ROI images, three features were extracted namely Chi-squared distance, Earth Mover’s Distance (EMD) and contrast measurement. Finally, the features will be tested using a paired t-test. This test was performed to analyse whether the features are statistically significant to classify the image into normal and abnormal classes. Based on the study done, Chi-squared distance and EMD features are statistical significance to be used in classification of normal and abnormal image. While, the contrast measurement is computed to be statistically insignificant for future classification work. |
|---|