Related Experiment Video
Updated: Jun 17, 2025
07:12
Establishment of a Simple and Effective Rat Model for Intraoperative Parathyroid Gland Imaging
Published on: August 17, 2022
3.4K
Intraoperative detection of parathyroid glands using artificial intelligence: optimizing medical image training with
Joon-Hyop Lee1, EunKyung Ku2, Yoo Seung Chung3
1Division of Endocrine Surgery, Department of Surgery, Samsung Medical Center, 81 Irwon-ro, Gangnam-gu, Seoul, Korea.
Surgical Endoscopy
|August 14, 2024
Summary
Artificial intelligence (AI) can effectively identify parathyroid glands during thyroidectomy. Augmentation methods significantly improve AI performance for intraoperative parathyroid gland detection, enhancing surgical safety.
Area of Science:
- Surgical Technology
- Artificial Intelligence in Medicine
- Endocrinology
Background:
- Postoperative hypoparathyroidism is a significant complication following thyroidectomy.
- Intraoperative imaging for artificial intelligence (AI) training is challenging.
- AI holds potential for intraoperative parathyroid gland detection via augmentation.
Purpose of the Study:
- To train an effective AI model for intraoperative parathyroid gland identification during thyroidectomy.
Main Methods:
- Collected video clips of parathyroid glands during thyroid lobectomy.
- Trained AI models using baseline, geometric transformation, and generative adversarial network (GAN)-based image inpainting datasets.
- Evaluated AI performance using average precision for parathyroid gland detection.
Main Results:
- Baseline AI model achieved 77% average precision.
- Geometric transformation (79%) and image inpainting (78.6%) augmentation improved performance.
- Image inpainting showed superior effectiveness (46%) in external validation with a different surgical approach.
Conclusions:
- The AI model is effective and generalizable for intraoperative parathyroid gland identification.
- Augmentation methods are crucial for enhancing AI performance in this application.
- Further studies are needed to compare AI performance with surgeon identification for clinical relevance.
