Screening Thyroid Tumor in Ultrasound Thyroid Images using Hidden Markov Model

Detection and segmentation of tumor regions in Ultra Sound (US) images is important for saving the human life. At present, computer aided methods are developed for detecting the tumor regions in US images. Hence, this paper develops computer aided tumor regions detection and segmentation in ultrasound thyroid images using Hidden Markov Model (HMM). This method computes the features from both benign and malignant thyroid images from the open access database. The computed features are trained and classified into benign and malignant using the improved hidden states with HMM approach. Then, region growing algorithm is applied on the malignant image to segment the tumor pixels in the classified malignant thyroid image. This proposed tumor detection and segmentation methodology is tested on both low resolution and high resolution thyroid image dataset for evaluating the proficiency of the developed model. The performance of the proposed method is well compared with other existing methods in terms of sensitivity, specificity and tumor segmentation accuracy.