This research work focuses on developing an advanced diagnostic method for thyroid nodules using ultrasonography images. The core idea revolves around the observation that the presence and amount of calcium flecks in thyroid nodules can indicate their severity, potentially leading to severe thyroid cancer. A novel technique, named Bilateral Mean Clustering Strategy (Bi-MCS), is proposed, combining the strengths of Fuzzy C mean and K-mean clustering approaches. This technique enhances color sense-based segmentation accuracy by precisely identifying the edges of thyroid nodules, crucial for determining their severity. This precise identification is achieved by analyzing the variations in pixel intensity associated with calcium flecks. Furthermore, this technique is incorporated into a deep convolutional neural network, specifically a modified ResNet101 structure, referred to as Bi-ResNet101. This DCNN framework is specifically designed to process and analyze the grayscale intensity profiles of ultrasonography images, focusing on the density of calcium flecks around thyroid states. The experimental analysis compares the efficiency of Bi-ResNet101 with other models like Resnet18, Resnet50, and standard Resnet101, demonstrating its superior capability in computing the density of calcium flecks and classifying different stages of thyroid nodules.