Harmonization of chest CT scans for different doses and reconstruction methods

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Computed tomography (CT) acquisition protocols vary significantly across vendors and reconstruction methods.
• These variations introduce noise and biases, compromising the reliability of densitometry measurements.
• Harmonization techniques are needed to standardize CT data for accurate quantitative analysis.

Purpose of the Study:

• To develop and validate a CT harmonization technique combining noise stabilization and autocalibration.
• To provide reliable densitometry measurements in heterogeneous CT acquisition protocols.
• To enable accurate comparisons between low-dose and high-dose CT acquisitions.

Main Methods:

• A novel method was developed to reduce spatially variant noise and biases in CT images.
• The technique statistically characterizes signal-to-noise relationships and employs autocalibration.
• Validation was performed using synthetic images and clinical chest CT scans from two vendors (Siemens and GE).

Main Results:

• The harmonization technique reduced nonhomogeneous noise effects by 25% (RMSE) without compromising image resolution.
• Average biases from dosage and reconstruction differences were reduced by up to 74.20%.
• Harmonized data enabled comparable or even higher statistical similarity between low-dose and high-dose acquisitions.

Conclusions:

• The proposed CT harmonization technique allows direct comparison of low-dose and high-dose acquisitions without a reference reconstruction.
• It is applicable to retrospective studies as it does not require pre-phantom calibration.
• This method is suitable for multicenter trials with diverse CT vendors and reconstruction techniques.