ABSTRACT
Nodules on the thyroid gland are common and increases with age, it can either be classified as benign or malignant. To determine whether a nodule is benign or malignant, several techniques are used, including percutaneous biopsy and ultrasound imaging. However, existing methodologies can lead to medical errors. The proposed DRNN and ultrasound images work together to enhance nodule characterization and decrease biopsies. This ensemble DRNN implementation correctly classified thyroid nodules based on TI-RADS(Thyroid Imaging Reporting & Data System classifications. The experimental system was built utilising the Thyroid Digital Image Database (TDID) to validate its functionality. The pre-trained ImageNet dataset was used to evaluate the model setup in addition to the same dataset being used for training and testing. Recall, and F1-score were used to determine the ensemble network model’s diagnostic performance.
