Related Experiment Video
Updated: Jul 16, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.9K
Aided diagnosis of thyroid nodules based on an all-optical diffraction neural network
Lingxiao Zhou1, Luchen Chang2, Jie Li1
1Nanophotonics Research Center, Institute of Microscale Optoelectronics, Shenzhen University, Shenzhen, China.
Quantitative Imaging in Medicine and Surgery
|September 15, 2023
Summary
This study introduces an all-optical diffraction neural network for thyroid nodule diagnosis. The optical neural network achieved high accuracy in classifying benign and malignant thyroid nodules using ultrasound images.
Area of Science:
- Medical imaging
- Artificial intelligence
- Endocrinology
Background:
- Thyroid cancer is the most common endocrine malignancy, often presenting as thyroid nodules.
- Ultrasound is the primary non-invasive tool for diagnosing thyroid nodules.
- Artificial intelligence (AI) is an emerging technology for diagnostic assistance.
Purpose of the Study:
- To propose and evaluate an all-optical diffraction neural network for auxiliary diagnosis of thyroid nodules.
- To assess the performance of optical neural networks in medical image processing.
Main Methods:
- Developed an all-optical diffraction neural network with 5 diffraction layers and 1 detection plane.
- Utilized ultrasound images from a dataset of patients undergoing thyroid nodule diagnosis and follow-up.
- Input images were processed using light at a 632.8 nm wavelength.
Main Results:
- Achieved 97.79% accuracy and 99.8% area under the curve for classifying benign versus malignant thyroid nodules.
- Demonstrated 84.92% accuracy in detecting the presence of thyroid nodules.
Conclusions:
- All-optical neural networks show significant potential for medical image processing applications.
- The performance of optical neural networks is comparable to existing image classification models.
