Anomaly Detection and Artifact Recovery in PET Attenuation-Correction Images Using the Likelihood Function

Area of Science:

• Medical Imaging Physics
• Radiological Sciences

Background:

• Dual modality PET/CT imaging relies on CT data for accurate PET attenuation correction.
• CT-to-PET attenuation conversion algorithms assume specific patient material composition.
• Contrast agents used in CT scans introduce anomalies, leading to attenuation artifacts in PET images.

Purpose of the Study:

• To develop and test a novel approach for correcting contrast agent-induced artifacts in PET/CT attenuation maps.
• To improve the accuracy of PET emission image reconstruction by addressing CT-based attenuation inaccuracies.

Main Methods:

• Proposed a likelihood-based algorithm to identify and correct contrast artifacts using inherent PET emission data.
• Exploited the characteristic of contrast artifacts appearing as elevated attenuation values in the CT-derived map.
• Evaluated the performance of the log-likelihood objective function within a detection/correction algorithm framework under varying noise conditions.

Main Results:

• The proposed algorithm successfully identified and corrected contrast-related attenuation inaccuracies in simulated data.
• Both the full correction algorithm and the log-likelihood component demonstrated robust performance with simulated datasets.
• The method leverages PET data to mitigate artifacts that can mimic diseased tissue in corrected PET images.

Conclusions:

• The developed likelihood-based algorithm shows promise for correcting contrast artifacts in PET/CT imaging.
• Further validation with patient data is necessary to fully ascertain the capabilities and clinical utility of the approach.
• Accurate attenuation correction is crucial for reliable PET/CT image interpretation and diagnosis.