A hybrid multi-panel image segmentation framework for improved medical image retrieval system

Area of Science:

• Medical Imaging
• Computer Vision
• Information Retrieval

Background:

• Multi-panel images are crucial for medical diagnostics, comprising ~50% of medical literature.
• These images consolidate diverse patient data (X-rays, MRIs, CT scans) for comprehensive diagnosis.
• Extracting sub-images from regular/irregular layouts is challenging for current medical image retrieval systems.

Purpose of the Study:

• To develop a novel hybrid framework for enhanced sub-image retrieval from multi-panel medical images.
• To address challenges in segmenting both regular and irregular multi-panel medical image layouts.
• To improve the accuracy and efficiency of medical image retrieval systems.

Main Methods:

• A hybrid framework combining image classification, computer vision, and image processing techniques.
• Utilized image projection profiles and morphological operations for precise segmentation.
• Developed efficient segmentation methods for regular and irregular multi-panel medical images.

Main Results:

• Achieved 90.50% accuracy in medical image type identification.
• Attained 91% accuracy in segmenting regular multi-panel images.
• Achieved 92% accuracy in segmenting irregular multi-panel images.

Conclusions:

• The proposed hybrid framework significantly enhances sub-image retrieval from diverse multi-panel medical images.
• Accurate and efficient segmentation across regular and irregular layouts improves medical image retrieval system performance.
• This approach holds substantial potential for advancing medical diagnostics and literature analysis.