PT  - JOURNAL ARTICLE
AU  - Joyita Dutta
AU  - Quanzheng Li
AU  - Kira Grogg
AU  - Xuping Zhu
AU  - Chuan Huang
AU  - Georges El Fakhri
TI  - PSF modeling and anatomical prior design for improved quantitative PET using a dedicated NeuroPET/CT
DP  - 2014 May 01
TA  - Journal of Nuclear Medicine
PG  - 369--369
VI  - 55
IP  - supplement 1
4099  - http://jnm.snmjournals.org/content/55/supplement_1/369.short
4100  - http://jnm.snmjournals.org/content/55/supplement_1/369.full
SO  - J Nucl Med2014 May 01; 55
AB  - 369 Objectives The quantitative accuracy of PET in neuroimaging applications is severely limited by partial volume effects due to the limited spatial resolution capabilities of PET scanners. Our objectives are 1) to develop a framework that enables use of anatomical prior information to correct for partial volume and improve quantitation and 2) to apply it to a high-resolution prototypic mobile brain PET/CT scanner, where gains can be maximized. Methods We developed a partial volume correction technique based on 1) deconvolution with the measured spatially variant point spread function (PSF) of the NeuroPET/CT scanner measured in the image domain and 2) minimization of the joint entropy between the PET image and an anatomical counterpart. The PSF was measured by placing 0.35 mm diameter point sources filled with [18F]FDG at different radial and axial positions inside the scanner bore, 25 cm in diameter and 21 cm in length. We adopted a novel Fourier-based approach for fast spatially varying deconvolution coupled with a Parzen window based approach for joint entropy minimization. For validation, we performed a 15 min PET acquisition on the Hoffman. A separately acquired T2-weighted MR image of the phantom was rigidly co-registered to the OSEM reconstructed PET image. To quantify contrast recovery, we computed the mean and standard deviation of standardized uptake values (SUVs) in the grey and white matter and calculated the grey-to-white matter SUV ratio. Results While the mean grey-to-white SUV ratio in the original PET image was 3.0, the MR-assisted deconvolution operation yielded a value of 3.9 (ground truth 4). The standard deviation in the grey matter was reduced by 7% while that in the white matter was reduced by 13%. Conclusions MR-assisted joint entropy based deconvolution enabled recovery of sharp edges and structural details in the PET images. Our experiments performed on a dedicated brain PET scanner show that this technique leads to substantial improvement in PET contrast and quantitation and is promising for a variety of neuroimaging applications.