Papillary Thyroid Carcinoma (PTC) is the most frequently occurring type of thyroid cancer. Diagnosing PTC using histopathological imaging as the gold standard in clinical practice remains challenging. The rise of Artificial Intelligence (AI) has accelerated the development of fast and accurate computer-aided pathology for PTC identification. Numerous studies have demonstrated that deep learning-based methods yield promising results; however, current approaches often overlook the nuclei, a key feature in PTC diagnosis. This study proposes a deep residual network with integrated StarDist nuclei segmentation to enhance the model’s ability to focus on critical cellular structures when making predictions. To assess its robustness, the model was trained on the Tharun & Thompson dataset and evaluated on three independent external datasets: an institutional hospital dataset, the Cancer Genome Atlas (TCGA) dataset, and the Nikiforov dataset. Our proposed method achieved 98.55% accuracy, 98.55% sensitivity, and 98.84% specificity in internal testing while also demonstrating an improved focus on nuclei-dense regions, aligning with the diagnostic approach used by pathologists. Our findings confirmed that incorporating nuclei segmentation significantly enhances classification performance and model interpretability. The proposed model enhances feature localization by integrating nucleus segmentation with Deep Residual Networks, yielding a practical and efficient solution for histopathological PTC analysis. © 2013 IEEE.