Medical image retrieval based on plaque appearance and image registration

Area of Science:

• Medical Imaging
• Computer Science
• Biomedical Engineering

Background:

• Hospitals generate vast amounts of medical images daily, necessitating efficient organization beyond manual annotation.
• Existing Content-Based Image Retrieval (CBIR) systems are often not optimized for the complexity of medical images.
• Medical image retrieval by content is an emerging field with significant application potential.

Purpose of the Study:

• To address the need for automated organization of medical images.
• To present a general framework for medical image retrieval based on plaque appearance.
• To highlight the specific challenges and requirements for retrieving medical images based on visual content.

Main Methods:

• Overview of existing research in medical image retrieval.
• Development of a general framework for content-based medical image retrieval.
• Focus on plaque appearance as a key feature for image organization.
• Incorporation of local, global, and contextual image features.
• Ensuring invariance to shape variations common in medical imaging.

Main Results:

• Demonstration of a framework for medical image retrieval based on plaque appearance.
• Application and evaluation of the framework using intravascular ultrasound images.
• Presentation of extensive results on retrieval performance for this specific modality.

Conclusions:

• Content-based retrieval, focusing on plaque appearance, offers a viable solution for medical image organization.
• Effective medical image retrieval requires sophisticated features that account for complex information and shape variance.
• Intravascular ultrasound imaging serves as a relevant case study for validating the proposed retrieval framework.