Analysis and Design of Cancer Detection and Segmentation Schemes

Abstract

With the advancement in thermal imaging technology, thermogram based methods are newlinebecoming increasingly popular for non-invasive and early detection of abnormalities. The newlinebreast thermograms can be used to detect location, physiological condition and vascular state newlineof anomalous breast tissues. The present work contributes in developing novel and efficient newlinemethodologies for breast anomaly detection using thermograms. In this context, cancer newlinedetection and segmentation schemes using different texture feature extraction, feature selection newlineand classification techniques are proposed. A comparative analysis to determine ability of newlinevarious texture features and classify the malignancy in breast tissues is also presented. The newlineproposed schemes are implemented over publicly available dataset of breast thermograms. newlineThe first scheme presents, a novel hybrid texture feature set and fractional derivative newlinefilter-based breast cancer detection model. Thermal images have intrinsic properties such as; newlinelow contrast, blurr edges and multiplicative noise. Therefore, breast thermal images are filtered newlineby Grumwald Letnikov fractional derivative based Sobel filter for enhancing the texture and newlinerectifying the noise. A novel hybrid feature set using statistical texture features is derived and newlinethe high dimensions of formed feature set are reduced by applying principal component analysis newlinemethod. The skills of Radial-basis-function support vector machine (RBF-SVM) for detection newlineare employed to anomaly detection in thermal images. The performance parameters of the newlineproposed scheme are determined using K-fold cross validation (K-FCV) method. The fractional newlineorder alfa (and#945;) offers an extra adaptability in overcoming the limitations of thermal imaging and newlineassists radiologist in prior breast cancer detection. The proposed scheme is more generalized newlinewhich can be used with different thermal image acquisition protocols and IoT based newlineapplications. newline