Abstract
1734
Objectives In kinetic analysis of dynamic FDG-PET data, accurate knowledge of the arterial input function (AIF) is required. The gold standard method, collection of arterial samples is invasive and laborious, and is usually avoided in practice. In this study, an alternative non-invasive method is presented to measure the image derived input function (IDIF) from FDG-PET images, using anatomical magnetic resonance (MR) images to delineate the carotid arteries for use in partial volume correction.
Methods Carotid arteries were segmented in MR images using a thresholding and region growing algorithm [1]. PET frames were co-registered to MR images using a 12 parameter affine model [2]. In order to account for limited spatial resolution of the PET scanners, single region voxel-wise [3] partial volume correction (PVC) method, which only requires segmented carotid arteries, was applied to PET images. Point spread function (PSF) with a FWHM of 6mm is used in the PVC. The mean intensity of the arterial voxels was computed for each PET frame to compute the IDIF. We have applied this method to FDG-PET and MRI brain data from 4 subjects. Arterial samples were collected to estimate the true AIF. Performance was assessed by comparing area under curves (AUCs) and influx constant, Ki, estimates obtained from the two AIFs.
Results Similar AUC values were seen for both AIFs with a mean absolute percentage difference of 12.37%. A good agreement in estimated Ki values were observed with a mean absolute percentage difference of 13.66%. In addition, t-test results showed no statistically significant differences between AUC and estimated Ki values across two AIFs.
Conclusions In conclusion, we have developed an IDIF extraction method which has comparable performance to the current gold standard method. Future work includes further test of the method with more human brain data.
Research Support This work was supported by an IMPACT studentship funded jointly by Siemens and the UCL Faculty of Engineering Sciences. K.E. is funded by a grant from EPSRC (EP/K005278/1). UCL/UCLH research is supported by the NIHR Biomedical Research Centres funding scheme.