Ultra-Dense Denoising Network: Application to Cardiac Catheter-Based X-Ray Procedures

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Reducing radiation dose in cardiac catheterization is crucial for patient safety.
• Low-dose X-ray imaging often suffers from increased noise and artifacts, potentially impacting clinical decisions.
• Effective denoising methods are needed to maintain image quality at reduced radiation levels.

Purpose of the Study:

• To propose and evaluate a novel convolutional neural network (CNN) framework, the Ultra-Dense Denoising Network (UDDN), for low-dose X-ray image denoising.
• To improve the diagnostic quality of cardiac X-ray images acquired with reduced radiation exposure.
• To provide a clinically adaptable solution for enhancing X-ray image quality in interventional cardiology.

Main Methods:

• Development of the Ultra-Dense Denoising Network (UDDN) featuring a novel residual block for enhanced feature extraction.
• Training the UDDN framework using normally-distributed noise and synthetic additive noise X-ray data.
• Clinical validation using real-world data from cardiac procedures and assessment by local SNR and cardiologist voting.

Main Results:

• The UDDN demonstrated statistically significant improvements in peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) compared to other methods on synthetic data.
• Clinical evaluation showed the UDDN outperformed comparative methods in local SNR and cardiologist assessments.
• The framework proved effective in denoising clinical X-ray data, indicating strong clinical adaptability.

Conclusions:

• The Ultra-Dense Denoising Network (UDDN) is a promising solution for low-dose X-ray image denoising in cardiac procedures.
• UDDN enables lower radiation doses without compromising essential image information, enhancing patient and healthcare professional safety.
• This AI-driven approach offers significant potential for improving diagnostic accuracy and clinical decision-making in interventional cardiology.