LightNet: A Novel Lightweight Convolutional Network for Brain Tumor Segmentation in Healthcare.

Area of Science:

• Medical imaging
• Artificial intelligence
• Neuroscience

Background:

• Neuroimaging-based tumor segmentation is crucial for diagnosing and treating brain diseases.
• Convolutional Neural Networks (CNNs) have improved brain tumor segmentation efficiency.
• Existing CNNs often increase computational cost with complex modules.

Purpose of the Study:

• To propose a simple and effective feed-forward CNN, LightNet, for efficient brain tumor segmentation.
• To reduce network parameters and computational complexity compared to existing methods.

Main Methods:

• LightNet utilizes multi-path and multi-level architecture with light operations, replacing traditional convolutions.
• A light channel attention module is incorporated in the up-sampling stage for enhanced feature extraction.
• The network was evaluated on the Multimodal Brain Tumor Segmentation Challenge (BraTS 2015) dataset.

Main Results:

• LightNet demonstrates comparable accuracy to other high-performing CNNs on the BraTS 2015 dataset.
• The proposed network significantly increases efficiency and segmentation performance.
• Redundancy and computational complexity are substantially reduced.

Conclusions:

• LightNet offers a high-performing solution for brain tumor segmentation with an optimal balance between efficiency and accuracy.
• The network's reduced computational demands enable better energy performance, particularly on mobile devices.
• This approach facilitates more accessible and efficient neuroimaging analysis for clinical applications.