Related Experiment Video
Updated: Jun 21, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
2.3K
Trans RCED-UNet3+: a hybrid CNN-transformer model for precise lung nodule segmentation
Sadaf Raza1, Razia Zia1,2, Irfan Ahmed Usmani3
1Department of Electronic Engineering, Sir Syed University of Engineering & Technology, Karachi, Pakistan.
Frontiers in Oncology
|November 3, 2025
Summary
This study introduces Trans RCED-UNet3+ for precise lung nodule segmentation in CT scans. The enhanced model improves accuracy for better lung cancer diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Accurate lung nodule segmentation in CT scans is critical for lung cancer diagnosis.
- Challenges include small nodule size and complex shapes, hindering precise segmentation.
Purpose of the Study:
- To present Trans RCED-UNet3+, an advanced framework for enhanced lung nodule segmentation.
- To improve the accuracy and efficiency of detecting lung nodules in medical images.
Main Methods:
- Utilized a transformer-based bottleneck for global context and long-range dependency capture.
- Incorporated residual connections for efficient feature flow and gradient preservation.
- Employed a hybrid Dice and Binary Cross-Entropy loss function to refine nodule boundary segmentation.
Main Results:
- Trans RCED-UNet3+ achieved a Dice score of 0.990 on the LIDC-IDRI dataset.
- This represents an improvement over the original RCED-UNet3+ model's score of 0.984.
Conclusions:
- The integration of convolutional and transformer architectures offers a robust approach for medical image segmentation.
- The proposed model enhances the detection of subtle and irregular lung nodule structures, supporting more accurate clinical diagnoses.
