ChatGPT-assisted deep learning model for thyroid nodule analysis: beyond artifical intelligence.

Area of Science:

• Artificial Intelligence in Medicine
• Medical Imaging Analysis
• Deep Learning for Diagnostics

Background:

• Thyroid nodules require accurate diagnosis, often relying on fine needle aspiration biopsy (FNAB) cytopathology.
• Deep learning models offer potential for automated analysis of medical images, including thyroid ultrasound.
• Integrating AI tools like ChatGPT can streamline the development of these complex models.

Purpose of the Study:

• To develop a deep learning model for thyroid nodule classification using ultrasound images.
• To leverage ChatGPT's capabilities in AI model development for medical image analysis.
• To establish a baseline using FNAB cytopathology for model validation.

Main Methods:

• A retrospective analysis of 1,061 patients' thyroid ultrasound images and FNAB results (2017-2022).
• Training a deep learning model on imaging characteristics and cytological features to identify thyroid pathologies.
• Utilizing ChatGPT for AI model development, including coding, preprocessing, optimization, and troubleshooting.

Main Results:

• The deep learning model achieved an accuracy of 0.81 (95% CI: 0.76-0.87) on the test set.
• High performance was observed in classifying benign nodules (F1-score: 0.86) and malignant nodules (F1-score: 0.87).
• The model demonstrated strong precision and recall across different thyroid nodule categories.

Conclusions:

• AI, particularly with ChatGPT's assistance, shows significant potential in developing robust deep learning models for medical image analysis.
• This approach can enhance the accuracy and efficiency of diagnosing thyroid nodules.
• Further research can explore broader applications of AI in diagnostic imaging.