DENSE-INception U-net for medical image segmentation.

Area of Science:

• Deep Learning
• Medical Image Analysis
• Computer Vision

Background:

• Convolutional Neural Networks (CNNs) are crucial for medical image segmentation.
• The U-net architecture is a popular choice for semantic segmentation tasks.
• Deeper networks can improve performance but face challenges like gradient vanishing.

Purpose of the Study:

• To propose a novel CNN architecture integrating Inception-Res and dense connections into U-net.
• To enhance medical image segmentation performance by addressing limitations of standard U-net.
• To develop a robust model for various medical imaging modalities.

Main Methods:

• Integrated Inception-Res blocks to increase network width.
• Incorporated Dense-Inception blocks for deep feature extraction without extra parameters.
• Utilized down-sampling and up-sampling blocks for efficient processing.

Main Results:

• Achieved an average Dice score of 0.9857 for lung segmentation.
• Reached an average Dice score of 0.9582 for blood vessel segmentation.
• Obtained an average Dice score of 0.9867 for brain tumor segmentation.

Conclusions:

• Combining Inception modules with dense connections in U-Net is effective for semantic medical image segmentation.
• The proposed model demonstrates state-of-the-art performance across multiple segmentation tasks.
• This approach offers a promising direction for advancing medical image analysis.