Study on diagnosis of thyroid nodules based on convolutional neural network

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Oncology

Background:

• Thyroid nodules are common, and accurate diagnosis of malignancy is crucial for appropriate patient management.
• Distinguishing benign from malignant thyroid nodules using ultrasound can be challenging, necessitating advanced diagnostic tools.

Purpose of the Study:

• To evaluate the performance of a convolutional neural network-based artificial intelligence (AI) algorithm in diagnosing the nature and classifying thyroid nodules.
• To compare the diagnostic efficacy of AI, sonographers, and a combined AI-sonographer approach.

Main Methods:

• Retrospective analysis of 105 patients with surgically or pathologically confirmed thyroid nodules.
• Evaluation of nodule properties (solid components, echogenicity, borders, A/T ratio, calcification) by sonographers and AI.
• Generation of receiver operating characteristic (ROC) curves to assess diagnostic performance.

Main Results:

• The AI algorithm demonstrated higher sensitivity (84.5%) and specificity (81.0%) than sonographers (80.7% sensitivity, 73.7% specificity) for diagnosing benign and malignant thyroid nodules.
• The combined AI and sonographer diagnosis achieved superior performance with 92.1% sensitivity, 86.3% specificity, and an area under the curve (AUC) of 0.910.

Conclusions:

• The combined AI and sonographer approach significantly improves diagnostic accuracy for thyroid nodules compared to either method alone.
• This integrated diagnostic strategy can help reduce unnecessary fine-needle aspiration biopsies and guide surgical decisions more effectively.