ABSTRACT
PURPOSE: To understand if unexplained signal artifacts in MRg-LITT proton resonance frequency- (PRF-) shift thermometry images are caused by air bubbles or hemorrhages, and to characterize their effects on temperature measurements. METHODS: Retrospective image data from an IRB-approved clinical trial of intracranial MRg-LITT were inspected for asymmetric distortions observed in phase data during ablations, which have been previously reported as likely hemorrhages. A total of eight patient cases were selected: seven with artifact occurrence and one without. Mathematical image models for air bubbles or hemorrhages were implemented to estimate the size of the air bubble or hemorrhage needed to explain the clinically observed phase artifacts. Correlations and Bland-Altman analyses were used to determine if an air bubble model or a hemorrhage model was better correlated to the clinical data. The model was used to inject bubbles into clean PRF phase data without artifacts to examine how temperature profile distortions change with slice orientation. The simulated air-bubble injected data were compared to clinical data containing artifacts to examine the bubbles' effects on temperature and thermal damage estimates. RESULTS: The model demonstrated that air bubbles up to approximately 1 cm in diameter could explain the clinically observed phase artifacts. The bubble model predicts that a hemorrhage would have to be 2.2 times as large as an air bubble in order to explain the same extent of phase distortion observed in clinical data. Air bubbles had 16% percent higher correlations to the clinical PRF phase data than hemorrhages, even after rescaling the hemorrhage phases to better match the data. The air bubble model also explains how the phase artifacts lead to both large positive and large negative temperature errors, up to ±100 °C, which could cascade to damage estimate errors of several millimeters. CONCLUSION: Results showed that the artifacts are likely caused by air bubbles rather than hemorrhages, which may be introduced before heating or appear during heating. Manufacturers and users of devices that rely upon PRF-shift thermometry should be aware these phase distortions from bubble artifacts can result in large temperature errors.
