Evaluation of yttrium-90 positron emission tomography dosimetry
Purpose: Radioembolization is a novel treatment which utilizes the liver's unique dual system blood supply to trap yttrium-90 (90Y) microspheres in microvasculature near liver tumors. Radioembolization dose planning and dosimetry are based on crude, inaccurate assumptions due to the lack of knowledge of patient specific 90Y microsphere distribution. In recent years, the very small 3.1867e-5 internal pair production decay branch of 90Y has been shown to allow for positron emission tomography (PET) imaging following radioembolization. This work explores the accuracy and limitation of 90Y PET imaging due to the extremely low signal to noise (SNR) ratio associated with 90Y and verifies the accuracy of using these PET images for 3-dimensional (3D) dosimetry. Material and Methods: PET acquisitions of a phantom containing 90Y filled cylindrical inserts were acquired to determine quantitative accuracy of the PET images to measure 90Y activity. Numerous reconstruction algorithms were used to determine the optimal protocol to balance image noise and accuracy. A GATE model of the PET scanner was used to evaluate the origin of prompt signal and random noise coincidence counts in these PET acquisitions. PET images were converted to dose maps using standard S-kernel convolution. Polymer gel dosimetry was used to validate the 3D dose map results. Furthermore, PET, with associated CT images, were used as input data into MC simulations to model dose rates surrounding patients for future patient release studies. A Siemens® Biograph 64 TruePoint PET/CT was used for all acquisitions and reconstructions. Results: The phantom study determined Siemens® OSEM-PSF algorithm, known as TrueX, with 2 iterations and 14 subsets had the optimal balance of noise and accuracy. Using this reconstruction algorithm, the PET images were found to accurately measure activity and calculated dose within 10% when 90Y concentration was above the minimum detectable concentrations (MDC) of 1 MBq/ml. However, this reconstruction algorithm was shown to have a positive bias in areas where concentration was below the MDC due to truncation of negative sinogram bin values caused by statistical noise in the random correction. Polymer gel dosimetry verified the accuracy of PET dose maps but also identified a limitation in cases of highly gradient distributions due to the PET spatial resolution spreading of measured activity. Additionally, external dose rates were found to be accurately predicted through use of 90Y PET/CT images as inputs into a MC simulation. Conclusion: Research in 90Y PET/CT has quickly been expanding over recent years as a feasible method to provide liver distribution of 90Y following radioembolization. This study demonstrates the accuracy and limitations of the use of these 90Y PET/CT images in patient specific qualitative dosimetry.
Hutchins, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our