Insights about cervical lymph nodes: Evaluating deep learning-based reconstruction for head and neck computed...

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Dual-energy computed tomography (CT) is crucial for head and neck cancer imaging.
• Image reconstruction algorithms significantly impact diagnostic accuracy.
• Evaluating novel reconstruction techniques is essential for optimizing image quality.

Purpose of the Study:

• To compare cervical lymph node image quality using deep learning-based image reconstruction (DLIR) versus traditional methods (FBP, ASiR-V) in head and neck cancer patients.
• To assess noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and sharpness of lymph nodes.
• To evaluate overall image quality, artifacts, and noise in selected regions.

Main Methods:

• Seventy head and neck cancer patients underwent contrast-enhanced dual-energy CT.
• Datasets were reconstructed using Filter Back Projection (FBP), hybrid Iterative Reconstruction (IR) with Adaptive Statistical IR (ASiR-V), and DLIR (low, medium, high).
• Image quality metrics (noise, SNR, CNR, sharpness) were quantitatively and qualitatively assessed by radiologists.

Main Results:

• DLIR significantly reduced noise (3.8-35.9%) and improved SNR (11.5-105.6%) and CNR (10.5-107.5%) compared to FBP and ASiR-V (p < 0.001).
• 0.625-mm DLIR images showed lower noise than 2.5-mm FBP/ASiR-V images.
• DLIR-medium and DLIR-high demonstrated sharper lymph node margins and vessels (p < 0.05), with superior overall image quality agreed upon by both readers.

Conclusions:

• DLIR-medium and DLIR-high reconstruction provide superior cervical lymph node image quality in head and neck CT.
• Enhanced image quality with DLIR supports potential use in dose-reduction protocols for thin-slice imaging.