Enhancing Ultrasound-Based Diagnosis of Unilateral Diaphragmatic Paralysis With a Visual Transformer-Based Model

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Diagnostic Tools

Background:

• Unilateral Diaphragmatic Paralysis (UDP) diagnosis can be challenging.
• Ultrasound (US) imaging offers a non-invasive method for diaphragm assessment.
• Current diagnostic methods may lack precision or efficiency.

Purpose of the Study:

• To develop and validate a novel AI-driven framework for UDP diagnosis using US images.
• To leverage deep learning, specifically Visual Transformer (ViT), for enhanced feature extraction.
• To improve diagnostic accuracy and efficiency in identifying UDP.

Main Methods:

• A pre-trained Visual Transformer (ViT) model was utilized for feature extraction from US images.
• A custom denoising image filter was integrated to enhance image quality.
• Extracted features were processed by an ensemble learning model for UDP classification.
• The framework was evaluated using stratified 5-fold cross-validation on data from 17 volunteers.

Main Results:

• The proposed framework achieved an average accuracy of 93.8% in diagnosing UDP.
• Performance surpassed existing state-of-the-art (SOTA) image classifiers.
• The method demonstrated robustness in capturing critical diagnostic features from US images.

Conclusions:

• The novel ViT-based framework offers a highly accurate and effective method for UDP diagnosis.
• This approach shows significant potential as a valuable diagnostic tool in medical imaging.
• The integration of deep learning with US imaging advances diagnostic capabilities for UDP.