Development, validation, and application of a generic image-based noise addition method for simulating reduced dose...

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Computed tomography (CT) aims to reduce patient radiation dose while preserving image quality.
• Existing simulation methods (image-based and projection-based) have limitations in realism or clinical practicality.
• Realistic simulation of reduced-dose CT is crucial for protocol optimization and dose reduction strategies.

Purpose of the Study:

• To develop and validate an image-based noise addition method for simulating reduced-dose CT images.
• To ensure the simulation method captures realistic noise attributes like texture and non-stationarity.
• To create a clinically practical tool for assessing radiation dose reduction in CT.

Main Methods:

• Developed an image-domain noise addition technique estimating noise power spectrum (NPS).
• Forward-projected images, added white noise proportional to attenuation, then back-projected and filtered.
• Validated using phantoms (Mercury, anthropomorphic) and patient data, comparing noise magnitude and texture (NPS-f_av).

Main Results:

• Simulated phantom images showed low noise magnitude errors (3.34-3.50%) and comparable NPS-f_av.
• Simulated patient images had a 4.61% average noise magnitude error, with visually similar noise texture.
• Clinical implementation proved practical, simplifying dose reduction estimation and enabling a 50% dose reduction in a multiple myeloma protocol.

Conclusions:

• The developed method successfully generates simulated CT images with realistic noise properties.
• It mimics noise characteristics of actual low-dose acquisitions without requiring raw projection data.
• This tool offers a practical approach for evaluating and optimizing CT radiation dose reduction.