Related Experiment Video
Updated: Sep 22, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Multiscale and Hierarchical Feature-Aggregation Network for Segmenting Medical Images
Nagaraj Yamanakkanavar1, Jae Young Choi2, Bumshik Lee1
1Department of Information and Communications Engineering, Chosun University, Gwangju 61452, Korea.
Sensors (Basel, Switzerland)
|May 20, 2022
Summary
This study introduces a novel deep learning model for medical image segmentation, enhancing accuracy in skin lesion detection. The architecture utilizes advanced feature aggregation and guided skip connections for superior performance.
Area of Science:
- Medical image analysis
- Computer vision
- Deep learning
Background:
- Accurate medical image segmentation is crucial for diagnosis and treatment planning.
- Existing methods often struggle with complex feature representation and information fusion.
Purpose of the Study:
- To develop an advanced encoder-decoder architecture for improved medical image segmentation.
- To enhance feature representation and fusion using novel aggregation modules.
Main Methods:
- Proposed an encoder-decoder architecture with wide and deep convolutional layers.
- Introduced multiscale feature aggregation (MFA) and hierarchical feature aggregation (HFA) modules.
- Employed MFA-based long residual connections and a guided block with multilevel convolution.
Main Results:
- Achieved superior segmentation performance compared to conventional methods.
- Demonstrated high accuracy (0.97 average score) on skin lesion segmentation datasets (ISIC-2018, PH2, UFBA-UESC).
- The proposed model effectively recovers spatial information and improves similarity to ground-truth maps.
Conclusions:
- The novel architecture with feature-aggregation modules and guided skip connections significantly enhances medical image segmentation accuracy.
- The model shows great potential for clinical applications, particularly in skin lesion analysis.
