Classification of Carotid Plaque with Jellyfish Sign Through Convolutional and Recurrent Neural Networks Utilizing...

Area of Science:

• Medical imaging analysis
• Cardiovascular diagnostics
• Artificial intelligence in healthcare

Background:

• Carotid artery plaque can present as elevated lesions.
• The Jellyfish sign, characterized by fluctuating plaque surfaces due to blood flow pulsation, is a notable feature.
• Early detection of the Jellyfish sign is crucial due to its association with cerebral infarction.

Purpose of the Study:

• To develop and validate an automated ultrasound video-based classification method for the Jellyfish sign.
• To leverage deep neural networks for accurate and efficient detection of this critical plaque characteristic.

Main Methods:

• Preprocessing carotid ultrasound videos to isolate vascular wall and plaque movements.
• Integrating preprocessed video data with plaque surface information.
• Utilizing a deep learning model combining convolutional and recurrent neural networks for classification.

Main Results:

• The proposed deep learning method successfully classified the Jellyfish sign in ultrasound videos.
• Validation was performed on data from 200 patients.
• Ablation studies confirmed the contribution of individual components within the deep learning model.

Conclusions:

• The developed deep neural network approach provides an effective means for classifying the Jellyfish sign from carotid ultrasound videos.
• This method holds potential for improving the early detection of plaques associated with cerebral infarction.