Segmentation of Medical Image Using Novel Dilated Ghost Deep Learning Model.

Area of Science:

• Computer Vision
• Medical Imaging Analysis
• Artificial Intelligence in Healthcare

Background:

• Image segmentation is crucial for computer-aided design but resource-intensive.
• Existing medical image segmentation software is limited.
• Extracting distinctive features requires specialized technical knowledge.

Purpose of the Study:

• To develop an effective Convolutional Neural Network (CNN) model for medical image segmentation.
• To improve the accuracy and efficiency of medical image analysis.
• To address the limitations of current medical image segmentation techniques.

Main Methods:

• A 13-layer CNN model incorporating dilated convolution and max-pooling for feature extraction.
• Utilizing a Ghost model to reduce feature duplication and complexity.
• Applying morphological techniques for pixel categorization and thickening.

Main Results:

• The proposed CNN model achieved 96.05% accuracy, 98.2% precision, and 95.78% recall.
• Demonstrated superior performance compared to traditional medical image segmentation models.
• Successfully segmented medical images, enabling acquisition of initial regions.

Conclusions:

• The developed segmentation strategy is effective for medical imaging.
• The CNN model offers a significant improvement over existing methods.
• This research provides a novel approach to medical image segmentation.