Encoder-Decoder Architecture for Ultrasound IMC Segmentation and cIMT Measurement

Area of Science:

• Medical Imaging
• Cardiovascular Disease Research
• Artificial Intelligence in Healthcare

Background:

• Cardiovascular diseases (CVDs) are a leading cause of global mortality.
• Intima-media thickness (IMT) measurement of the common carotid artery (CCA) is crucial for early CVD diagnosis.
• Current computer vision methods for CCA image analysis are limited by complexity and data scarcity.

Purpose of the Study:

• To develop a deep learning approach for semantic segmentation of the intima-media complex (IMC).
• To enable accurate calculation of carotid intima-media thickness (cIMT) for early CVD detection.
• To address the challenge of limited datasets in medical image analysis.

Main Methods:

• An encoder-decoder deep learning architecture was employed.
• Multi-image inputs were utilized to enhance model learning from diverse features.
• Semantic segmentation was applied to the IMC for precise measurement.

Main Results:

• The proposed deep learning model achieved effective IMC segmentation.
• The system accurately computed IMT thickness.
• Evaluation using image segmentation metrics demonstrated the architecture's effectiveness.

Conclusions:

• The developed model offers a robust and fully automated solution for cIMT measurement.
• This approach advances early CVD diagnosis through automated image analysis.
• The multi-image input strategy effectively mitigates dataset limitations.