[Research progress on deep learning-based computer-aided diagnosis of thyroid nodules using ultrasound imaging]

Area of Science:

• Endocrinology
• Medical Imaging
• Artificial Intelligence

Background:

• Thyroid nodules are common endocrine disorders requiring early detection to prevent thyroid cancer.
• Heterogeneous nodule morphology and boundaries present diagnostic challenges, increasing misdiagnosis risks with traditional methods.
• Computer-aided diagnosis (CAD) using deep learning offers potential for improved thyroid nodule analysis.

Purpose of the Study:

• To review recent advancements in deep learning-based CAD for thyroid nodule ultrasound image analysis.
• To focus on applications in image preprocessing, segmentation, and classification.
• To analyze current techniques' advantages and limitations and discuss future directions.

Main Methods:

• Review of current literature on deep learning algorithms applied to thyroid nodule ultrasound images.
• Analysis of methods for image preprocessing, segmentation, and classification.
• Evaluation of the performance, advantages, and limitations of existing deep learning models.

Main Results:

• Deep learning-based CAD methods demonstrate significant potential in analyzing thyroid nodule ultrasound images.
• These methods show promise in improving the accuracy of nodule identification and classification.
• Current techniques offer advantages but also have limitations that require further research.

Conclusions:

• Deep learning holds substantial potential for enhancing thyroid nodule diagnosis.
• Further research is needed to overcome limitations and optimize clinical application pathways.
• This review provides a foundation for future clinical integration of AI in thyroid nodule management.