Automatic Recognition and Classification System of Thyroid Nodules in CT Images Based on CNN

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Thyroid nodules are common, with increasing incidence over recent decades.
• X-ray computed tomography (CT) is crucial for thyroid disease diagnosis.
• Traditional machine learning struggles with complex thyroid CT images due to artifacts.

Purpose of the Study:

• To develop an end-to-end system for automatic recognition and classification of thyroid nodules using deep learning.
• To improve the accuracy and efficiency of diagnosing thyroid nodule malignancy.

Main Methods:

• An improved Eff-Unet segmentation network was utilized to segment thyroid nodules as regions of interest (ROI).
• A novel CNN-F network, featuring low-level and high-level feature fusion, was proposed for classifying nodules as benign or malignant.
• The system integrates segmentation and classification modules for automated nodule analysis.

Main Results:

• The segmentation Intersection over Union (IOU) reached 0.855 on the test set.
• The automatic classification accuracy for thyroid nodules achieved 85.92%.
• The developed system demonstrated excellent performance in diagnosing thyroid diseases.

Conclusions:

• The proposed end-to-end deep learning system effectively automates thyroid nodule recognition and classification.
• The system shows significant potential for improving diagnostic accuracy in thyroid imaging.
• This approach overcomes limitations of traditional methods in processing complex CT images.