Purpose: To validate the feasibility of thyroid nodule detection using attenuation value based on non-enhancement computed tomography (CT) images. Materials and Methods: One hundred and thirty-four transverse CT images with nodules from 58 inpatients and 128 normal images from 55 outpatients (healthy controls) were enrolled in this study. The inpatients with thyroid nodules (50 malignant, 84 benign) underwent nodule excision operation and final diagnoses were confirmed by histopathology. Thyroid regions of interest (ROIs) from axial CT images were delineated manually by a radiologist. The CT values of every thyroid pixels were extracted from the DICOM images. Median and average filter were applied to reduce image noise. Attenuation values of every 2∗2 matrix were compared to the high and low density thresholding to identify the presence of the low density area such as cyst and necrosis or the high density area like calcification. The parameters, thresholding and filter type, were optimized according to accuracy and sensitivity. To evaluate the performance of the proposed method, accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were considered. Result: The experimental results demonstrate that our proposed method offers exceptional accuracy (ACC=0.8511), sensitivity (SEN=0.8060), specification (SPC=0.8984), positive predictive value (PPV=0.8926) and negative predictive value (NPV=0.8156) respectively. Conclusion: Our study provides a practical strategy for addressing thyroid nodule detection. The proposed and deployed thresholding optimization approach could serve as computer-aided diagnosis method in clinical application.