Lesion insertion in the projection domain: Methods and initial results.

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Task-based image quality assessment in CT requires realistic patient images with known diagnostic truth.
• Hybrid images combining patient scans with inserted lesions are effective but conventional methods lack realism.
• Image-domain hybrid image generation fails to account for scan and reconstruction parameter impacts on lesion appearance.

Purpose of the Study:

• To develop and validate a projection-domain approach for generating realistic hybrid CT images with inserted lesions.
• To overcome limitations of image-domain methods in reflecting scan and reconstruction parameter effects.
• To enhance the realism and diagnostic utility of hybrid CT images for quality assessment.

Main Methods:

• Lesions were segmented and forward projected to create lesion projections.
• Lesion projections were inserted into patient projections from commercial CT data.
• Combined projections were reconstructed into hybrid images on a commercial CT scanner.
• Validation involved phantom studies for geometric accuracy and radiologist assessment for lesion realism.

Main Results:

• Forward-projection geometry accurately reproduced phantom image characteristics (CT number, resolution).
• Various lesion types were realistically inserted, indistinguishable from original lesions by expert radiologists.
• Lesion appearance in reconstructed hybrid images varied with different scan and reconstruction parameters.

Conclusions:

• A robust framework for projection-domain lesion insertion in CT images was established.
• The method accurately models the influence of scan and reconstruction parameters on lesion appearance.
• This projection-domain approach offers greater anatomical complexity and realism than previous methods.