Deep Learning-Based Iodine Contrast Augmentation for Suboptimally Enhanced CT Pulmonary Angiography: Implications for...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Cardiovascular Imaging

Background:

• Suboptimal contrast enhancement in CT pulmonary angiography (CTPA) can hinder accurate pulmonary embolism (PE) detection.
• Developing methods to improve image quality in these challenging cases is crucial for patient diagnosis.

Purpose of the Study:

• To evaluate the effectiveness of a deep learning-based iodine contrast augmentation (DLCA) algorithm.
• To assess DLCA's impact on image quality and diagnostic performance for PE detection in suboptimal CTPA scans.

Main Methods:

• Retrospective analysis of 103 suboptimal CTPA cases.
• Comparison of image quality metrics (attenuation, noise, SNR, CNR) between original and DLCA-processed images.
• Assessment of diagnostic performance for PE detection (per segment) with and without DLCA.

Main Results:

• DLCA significantly improved pulmonary artery opacification (+57.7%) and reduced noise (-56.7%).
• Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) showed substantial increases (p < 0.001).
• Diagnostic accuracy for PE detection improved across overall, central, and peripheral regions (AUCs increased significantly, p ≤ 0.003).

Conclusions:

• DLCA processing markedly enhances image quality in suboptimal CTPA.
• The algorithm significantly boosts diagnostic accuracy for pulmonary embolism detection.
• DLCA offers a valuable strategy for optimizing CTPA scans without additional contrast media or radiation dose.