Analysis of Breast Thermal Images Using Anisotropic Diffusion Filter Based Modified Level Sets and Efficient Fractal Algorithm
| Autor: | S. Prabha, C. M. Sujatha, S. S. Suganthi |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Ground truth
020205 medical informatics Binary image Diffusion filter Feature extraction 02 engineering and technology Fractal dimension 030218 nuclear medicine & medical imaging 03 medical and health sciences 0302 clinical medicine Level set Fractal Feature (computer vision) 0202 electrical engineering electronic engineering information engineering skin and connective tissue diseases Algorithm Mathematics |
| Zdroj: | Communications in Computer and Information Science ISBN: 9789811090585 |
| DOI: | 10.1007/978-981-10-9059-2_2 |
| Popis: | Asymmetry analysis of thermal images plays a prominent role in automated diagnosis of normal and abnormal breast tissues. Accurate segmentation of breast tissues further enhances the analysis. In this work, asymmetry analysis of breast thermal images is carried out using efficient fractal algorithm. Prior to feature extraction, the breast tissues are delineated by integrating anisotropy diffusion filter into modified level set framework. In order to validate the segmented results against ground truth images, Overlap measures are estimated. Breast regions such as left and right are separated from the segmented tissues. A set of binary images are generated by subjecting the separated breasts to thresholded binary decomposition (TBD) method. The features such as fractal dimensions, mean gray level and pixel count are computed from the resultant binary images. The segmented results had shown high degree of correlation against the ground truth images. Accuracy of the segmented results is found to be 98%. Asymmetry analysis using TBD method shows that the features are capable of discriminating the change in structural information which is caused by varied metabolic conditions. Finally, mean gray level feature show 11% improvement in demarcating normal and abnormal breast tissues. Therefore, the proposed method appears to be efficient in computer diagnosis of early breast abnormalities using infrared images. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|