Accurate thyroid cancer classification using a novel binary pattern driven local discrete cosine transform descriptor

Abstract

Background: :Thyroid cancer often manifests as small nodules in the ultrasound images that are difficult to manually classify into different classes. The existing Computer-Aided Diagnosis (CAD) systems do not achieve high levels of classification accuracy. Hence, there a is need for developing better CAD systems in this domain.Methods:A CAD system typically consists of feature extraction using image descriptors and classification using machine learning techniques. It has been shown in recent works that thyroid texture is an important feature for segregating the thyroid ultrasound images into distinct classes. Based upon our conjecture that this is the most significant feature, we propose a novel image descriptor that excels in extracting texture. Our new descriptor intelligently integrates two popular feature extracting methods

Local Discrete Cosine Transform and Improved Local Binary Pattern. The final classification is carried out using a non-linear Support Vector Machine.Results:The proposed CAD system is evaluated on the only two publicly available thyroid cancer datasets, namely TDID and AUITD. The evaluation is conducted in two stages (Benign/Malignant and different cancer types). For Stage I classification, our proposed model demonstrates exceptional performance of nearly 100% on TDID and 97% on AUITD. In Stage II classification, the proposed model again attains excellent classification of close to 100% on TDID and 99% on AUITD.Conclusion:Our proposed thyroid CAD system delivers superior performance and outperforms all the existing state-of-the-art studies. This will have a big impact in assisting doctors for performing more accurate diagnoses. © 2026 Elsevier B.V. All rights are reserved, including those for text and data mining, AI training, and similar technologies.