Related Experiment Video
Updated: Nov 22, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.1K
ResBCDU-Net: A Deep Learning Framework for Lung CT Image Segmentation
Yeganeh Jalali1, Mansoor Fateh1, Mohsen Rezvani1
1Faculty of Computer Engineering, Shahrood University of Technology, Shahrood 3619995161, Iran.
Sensors (Basel, Switzerland)
|January 6, 2021
Summary
This study introduces a novel deep learning model for automatic lung CT image segmentation, achieving high accuracy. The Res BCDU-Net significantly improves segmentation performance for lung cancer detection applications.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Lung CT image segmentation is crucial for lung cancer detection but faces challenges like similar image densities and scanner variations.
- Current semi-automatic methods often lack accuracy and have high false-positive rates due to human factors.
Purpose of the Study:
- To propose a deep neural network architecture for automatic lung CT image segmentation.
- To enhance segmentation accuracy and reduce false positives in medical imaging.
Main Methods:
- A modified U-Net architecture (Res BCDU-Net) was developed, replacing the encoder with a pre-trained ResNet-34.
- Bidirectional Convolutional Long Short-term Memory (BConvLSTM) was used as an advanced integrator module.
- Extensive preprocessing and morphological operations were applied to CT images and ground truths.
Main Results:
- The proposed Res BCDU-Net achieved a high Dice coefficient of 97.31% on the LIDC-IDRI lung CT image database.
- The method demonstrated effectiveness in automatic lung CT image segmentation.
Conclusions:
- The developed deep neural network architecture significantly improves automatic lung CT image segmentation.
- The Res BCDU-Net offers a promising solution for accurate and efficient lung cancer detection support.
