Improved correction techniques to compensate for partial volume and spill-in effects in PET

A major barrier to accurate tumour quantification and detectability in positron emission tomography (PET) is partial volume effect (PVE). PVE could either be spill-in effect or spill-out effect, leading to overestimation and underestimation (or loss) of tumour activity respectively. Many correction techniques are being employed to restore the real tumour activity in a process known as resolution recovery. However, past studies have shown that most of these correction techniques only correct for one of these two effects, leaving the other uncorrected for. Therefore, this study is aimed at developing and evaluating new correction techniques which are capable of compensating for both effects. Digital XCAT2 phantom with an [
¹⁸
F]-Fluorodeoxyglucose (FDG) distribution was used to simulate the pelvic bed, consisting of a hot bladder, and surrounded by 6 lesions (L1-L6) with diameters ranging from 6 - 20 mm, placed at different distances from the bladder. Analytical simulation and reconstruction was done with Software for Tomographic Image Reconstruction (STIR) using simulated Siemens mMR scanner template. The reconstructed images were corrected for partial volume effect by using 2 existing techniques: Point spread function (PSF) reconstruction and background correction (BC) method; as well as 2 newly-developed techniques: hybrid kernelized method (HK) and improved local projection (iLP) method. To estimate the quantification and noise properties of these correction techniques, we carried out region of Interest (ROI) analysis using recovery coefficient (RC) and contrast-to-noise ratio (CNR). The result shows that OSEM and PSF could not correct for the spill-in effect from the bladder as lesions close to the bladder have an overestimated SUV
_max
of up 40%. The spill of activity from the bladder to the surroundings causes the background uptake to increase, thereby lowering the CNR of both OSEM and PSF. iLP and HK have steady RC for all lesions with an increased CNR. PSF is a robust correction method for lesions bigger than 8 mm and those farther away from the bladder. Hence, combining the new correction methods with PSF might give a more robust result. However, there is need to further validate these techniques with acquired phantom and patient datasets.