Initial experience of deep learning reconstruction algorithm in lung kernel: clinical usefulness for lung nodules at...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• TrueFidelity (TF) is a deep learning algorithm for image reconstruction.
• TF is now available in a lung kernel (TF Lung) for CT scans.
• Ultra-low-dose CT (ULD CT) requires advanced reconstruction techniques to maintain image quality.

Purpose of the Study:

• To evaluate the performance of TF Lung in ULD CT.
• To assess image quality and nodule sharpness using TF Lung.
• This is the first study to examine TF Lung in ULD CT.

Main Methods:

• 68 patients underwent ULD CT and contrast-enhanced CT (CE CT).
• ULD CT images were reconstructed using Adaptive statistical iterative reconstruction-V (AR50) and TF in standard and lung kernels.
• Image noise, signal-to-noise ratio, contrast-to-noise ratio, and nodule sharpness (full-width half-maximum) were objectively measured.

Main Results:

• TF Lung significantly reduced image noise compared to AR50 Lung (p < 0.001).
• TF Lung at low and medium strengths demonstrated nodule sharpness comparable to the reference standard (CE AR50 Lung).
• Malignancy-related imaging features in pathologically confirmed malignant nodules were evaluated.

Conclusions:

• TF Lung improves image quality in ULD CT by reducing noise.
• TF Lung preserves nodule sharpness, crucial for accurate diagnosis.
• TF Lung shows potential to enhance diagnostic accuracy in ULD CT for lung nodule evaluation.