Image segmentation method for boundary detection of breast thermography using random walkers

• Main Author: Moghbel, Mehrdad
• Format: Thesis
• Language: English
• Published: 2013
• Online Access: http://psasir.upm.edu.my/id/eprint/41504/1/ITMA%202013%203R.pdf; http://psasir.upm.edu.my/id/eprint/41504/
• Institution: Universiti Putra Malaysia

In breast thermography diagnostic, proper detection and segmentation of the areola area as well as detection of breast boundaries present the biggest challenge. As the boundaries of breasts especially in the upper quadrants are usually not present, this produces a great deal of challenge to segment breasts automatically resulting in the majority of the segmentation work done by operator. Although almost half of all breast cancers occur in the upper outer region of the breast known as the tail of the breast, most of the segmentation methods cannot segment the upper outer region of the breast with the adequate accuracy. Image segmentation approaches are usually based on the identification of characteristics or features of the object and leveraging on them to achieve a proper segmentation. In breast thermography the lack of defined edges on the upper boundaries of the breast and the fact that breasts have different shape, size and characteristics even between breasts of a single individual, makes segmentation of breasts a difficult task for most segmentation methods. In this thesis, a new framework for segmentation of breast and the areola is introduced and discussed. Unlike other segmentation methods, random walkers showed great tolerance for irregular heat patterns present on the image and in most cases the segmented images corresponds perfectly with the anatomical shape of the breasts. The random walkers was the only segmentation method in the literature capable of segmenting the axillary region of the breast. All images used for this study were captured by state of the art forward looking infrared (FLIR) thermal cameras and have good resolution and sensitivity. The developed algorithm needs no human intervention until the final result is displayed to the user, if the user is not satisfied with the segmentation results he/ she can appoint new seeds interactively to fine tune the segmentation. The performance of the proposed method was evaluated by a board of three professional radiologists and the final decision was based on the majority agreement. The segmentation was based on constant parameters among all images used in the study; these standard segmentation parameters achieved acceptable results in most cases. Nevertheless the proposed method was able to surpass the highest accuracy reported within the literature. Use of interactive segmentation can further enhance these results dramatically as all the standard images that were not segmented correctly by the automatically method were correctly segmented after the utilization of the interactive mode.