<p>Endocrine cancer is thyroid cancer, and its prevalence has been steadily rising worldwide. A nodule near the thyroid gland in the neck is the earliest sign of thyroid cancer. One of the most reliable and popular approaches for finding thyroid nodules is ultrasonography. However, it takes time and creates difficulties for the professionals to evaluate all of the slide pictures. The most commonly used automated prediction model for detecting thyroid cancer is a different machine learning algorithm. However, attaining accurate prediction with lesser error probability is quite difficult using the existing prediction model. To address this issue, Channel boosted-Convolutional Neural Network (CB-CNN) is developed to identify thyroid cancer. Pre-processing of source ultrasound image is done using adaptive median filter and dualistic sub image histogram equalization for improving the resolution of input image. The adaptive median filter is employed to remove the noise from the original image. Dualistic sub image histogram equalization is used to enhance the image’s contrast level for further processing. The pre-processed image is segmented by using SegNet in order to identify the region of the tumour. After that, the segmented image is classified using CBCNN for predicting thyroid cancer. Simulation analysis reported, the proposed model reached 96% of accuracy, 4% error, 94% precision and 93% specificity. As a result, the proposed strategy performs better than other current strategies, such as Deep Convolutional Neural Network (DCNN), ResNet101, InceptionV3 and VGG19. Through this study it is preferred that proposed model is the best option for identifying thyroid cancer with better accuracy.</p>