Related Experiment Video
Updated: Aug 17, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.9K
A hierarchical GAN method with ensemble CNN for accurate nodule detection.
Seyed Reza Rezaei1, Abbas Ahmadi2
1Department of Industrial Engineering and Management Systems, Amirkabir University of Technology, Tehran, Iran.
International Journal of Computer Assisted Radiology and Surgery
|December 15, 2022
Summary
This study introduces a novel generative adversarial network (GAN) model for improved lung nodule detection. The approach enhances early lung cancer diagnosis by significantly reducing errors in lung segmentation and nodule localization.
Area of Science:
- Medical imaging analysis
- Artificial intelligence in healthcare
- Radiology and diagnostic imaging
Background:
- Early lung cancer detection is critical for patient survival but challenging due to nodule ambiguity.
- Automated tools are essential for radiologists to improve diagnostic accuracy.
Purpose of the Study:
- To develop an advanced generative adversarial network (GAN) model for accurate lung nodule detection.
- To improve early lung cancer diagnosis through enhanced nodule identification and segmentation.
Main Methods:
- A two-step GAN model was developed, incorporating lung segmentation (U-net generator with focal loss) and nodule localization (Mask R-CNN generator with a novel loss function).
- Discriminator networks utilized an ensemble of convolutional neural networks (ECNNs) for robust feature extraction and decision-making.
Main Results:
- The proposed model demonstrated significant error reduction on the LUNA dataset, achieving approximately 35% less error in lung segmentation and 16% less error in nodule localization compared to state-of-the-art methods.
- Experiments validated the model's effectiveness in improving diagnostic accuracy for lung nodules.
Conclusions:
- The novel approach employs dual loss functions for generators and ECNNs for discriminators, enhancing overall performance.
- This method offers a promising advancement in automated lung nodule detection, aiding early cancer diagnosis.
