Noise Texture Deviation: A Measure for Quantifying Artifacts in Computed Tomography Images With Iterative...

Area of Science:

• Medical Imaging
• Radiology
• Image Reconstruction

Background:

• Iterative reconstruction (IR) algorithms in computed tomography (CT) reduce image noise but can introduce specific artifacts.
• Quantifying these IR-specific artifacts is crucial for accurate image interpretation.
• Advanced modeled IR (ADMIRE) and sinogram-affirmed IR (SAFIRE) are two prominent IR techniques.

Purpose of the Study:

• Introduce noise texture deviation as a quantitative parameter for evaluating IR-specific artifacts in CT images.
• Assess whether ADMIRE reduces IR-specific artifacts more effectively than SAFIRE.
• Evaluate artifact reduction in ex vivo liver phantoms and in vivo patient studies with hypodense liver lesions.

Main Methods:

• An abdominal phantom and 40 patients with hypodense liver lesions were studied.
• Images were reconstructed using filtered back projection (FBP), SAFIRE, and ADMIRE.
• Noise texture deviation and noise power spectra were calculated; image noise was measured. Subjective artifact evaluation and lesion conspicuity/confidence ratings were performed by blinded readers.

Main Results:

• Noise texture deviation was significantly reduced in ADMIRE compared to SAFIRE (P < 0.001) and correlated strongly with subjective artifact ratings (r = 0.88).
• IR-specific artifacts were significantly reduced with ADMIRE compared to SAFIRE at strength levels 3 or greater, both ex vivo and in vivo (P < 0.001).
• No significant differences in lesion conspicuity or confidence were found between SAFIRE and ADMIRE, though both were superior to FBP.

Conclusions:

• Noise texture deviation serves as a reliable quantitative measure for IR-specific artifacts in CT.
• ADMIRE demonstrates superior performance in reducing IR-specific artifacts compared to SAFIRE in liver imaging.
• The quantitative measure supports the clinical utility of ADMIRE for improved CT image quality.