Related Experiment Video
Updated: Aug 4, 2025
07:53
Author Spotlight: Advancing 3D Modeling for Enhanced Diagnosis and Treatment of Pulmonary Nodules in Early-Stage Lung Cancer
Published on: October 13, 2023
1.5K
A Novel Artificial Intelligence Based Denoising Method for Ultra-Low Dose CT Used for Lung Cancer Screening
Larisa Gorenstein1, Amir Onn2, Michael Green3
1Department of Diagnostic Radiology, Sheba Medical Center, Tel Hashomer, Israel; Diagnostic Radiology, Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel.
Academic Radiology
|April 5, 2023
Summary
Ultra-low-dose (ULD) computed tomography with AI-based denoising (dULD) significantly reduces radiation exposure in lung cancer screening. This AI method maintains diagnostic accuracy for nodules and other findings, improving upon standard ULD scans.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Lung cancer screening aims to detect malignancies early.
- Computed tomography (CT) is a key imaging modality for lung cancer screening.
- Radiation dose is a concern in screening programs.
Purpose of the Study:
- To evaluate ultra-low-dose (ULD) CT for lung cancer screening.
- To assess a novel artificial intelligence (AI)-based denoising method (dULD) for ULD CT.
- To compare the diagnostic performance of ULD and dULD with standard low-dose (LD) CT.
Main Methods:
- Prospective study of 123 patients undergoing LD and ULD CT scans.
- A fully convolutional network trained with perceptual loss was used for AI-based denoising (dULD).
- Two independent readers reviewed all image sets.
Main Results:
- ULD CT reduced radiation dose by an average of 76%.
- dULD showed improved performance over ULD, with a lower negative likelihood ratio for readers.
- Diagnostic accuracy for actionable pulmonary nodules and coronary artery calcifications (CAC) was maintained or improved with dULD.
Conclusions:
- AI-based denoising enables substantial radiation dose reduction in CT screening.
- The dULD method allows for significant dose reduction without compromising the detection of critical findings like pulmonary nodules or aortic aneurysms.
- dULD demonstrates high sensitivity and accuracy, comparable to standard LD CT.
