Simulation study of quantitative precision of the PET/X dedicated breast PET scanner

Abstract

The goal for positron emission tomography (PET)/X is measuring changes in radiotracer uptake for early assessment of response to breast cancer therapy. Upper bounds for detecting such changes were investigated using simulation and two image reconstruction algorithms customized to the PET/X rectangular geometry. Analytical reconstruction was used to study spatial resolution, comparing results with the distance of the closest approach (DCA) resolution surrogate that is independent of the reconstruction method. An iterative reconstruction algorithm was used to characterize contrast recovery in small targets. Resolution averaged [Formula: see text] full width at half maximum when using depth-of-interaction (DOI) information. Without DOI, resolution ranged from [Formula: see text] to [Formula: see text] for scanner crystal thickness between 5 and 15 mm. The DCA resolution surrogate was highly correlated to image-based FWHM. Receiver-operating characteristic analysis showed specificity and sensitivity over 95% for detecting contrast change from 5:1 to 4:1 (area under curve [Formula: see text]). For PET/X parameters modeled here, the ability to measure contrast changes benefited from higher photon absorption efficiency of thicker crystals while being largely unaffected by degraded resolution obtained with thicker crystals; DOI provided marginal improvements. These results assumed perfect data corrections and other idealizations, and thus represent an upper bound for detecting changes in small lesion radiotracer uptake of clinical interest using the PET/X system.

Keywords: breast positron emission tomography; imaging biomarker; quantitative accuracy; response to therapy.

Fig. 1

Schematic illustration of the PET/X scanner that consisted of four planar detectors. The rectangular FOV measured 200-mm wide by 80-mm high by 150-mm deep. Note that the axial view ($x-y$ plane) in PET/X corresponded to a coronal view of the patient.

Fig. 2

Orthogonal views of the breast phantom with 10 10-mm diameter hot spheres (red). Breast background data were taken from distributed spheres of the same size (shown in green). The phantom was voxelized on a 1-mm grid. The image slice shown in each view is indicated by the crosshairs in the other two views. Note that not all spheres lie in the cross-sections shown here. Also note that the axial view (top left) is along the PET/X scanner axis that will be along the patient in anterior–posterior direction, and thus corresponds to a coronal view in conventional tomographic orientations.

Fig. 3

Scatterplot showing the correlation between the image spatial resolution using PFDRX reconstruction (vertical axis) and $d_{\text{FWHM}}$ derived from DCA method (horizontal axis). Each data point represents one of the 10 point sources in the phantom for a given detector crystal thickness and LOR estimate method. The phantom consisted of analytical point sources in a uniform background.

Fig. 4

Scanner efficiency as a function of crystal thickness for the rectangular geometry of Fig. 1. LSO was used as the scintillator material. Events that were scattered in the object and events that were scattered in the crystals and eventually escaped were both rejected by an energy threshold of 510 keV.

Fig. 5

Orthogonal views of sample reconstructed images using the iterative algorithm (top row is PET/X axial cross-section at the anterior–posterior center, cf. Fig. 2). (a)–(c) The LOR estimation methods were first vertex, 3D-, and 2D-COM. All images shared the same gray scale. Sphere diameter was 5 mm and crystal thickness was 10 mm. Image voxel size was 1 mm. Iteration number was 15. No Gaussian smoothing was applied in this image.

Fig. 6

(a) RCs for various sphere sizes with crystal thickness of 10 mm. Both ${\mathrm{RC}}_{\max }$ and ${\mathrm{RC}}_{\text{mean}}$ were plotted. (b) RCs with various crystal thicknesses for both lesion sizes of 10 and 5 mm. Only ${\mathrm{RC}}_{\text{mean}}$ values were shown. (a) and (b) All three LOR estimate methods were included. The error bars represented ERMSE. Each image was treated with a postreconstruction Gaussian smoothing with $\text{sigma}=0.8\mathrm{mm}$ as mentioned previously.

Fig. 7

(a) Sensitivity as a function of reduction in TBR from a pretherapy value of 5:1. Each curve corresponds to one of four specificity values. (b) ROC curve for four TBR reductions: $-20\%$, $-15\%$, $-10\%$, and $-5\%$. Both plots were computed with the data set of sphere diameter of 5 mm and crystal thickness of 10 mm. The LOR estimation method was 2D-COM (without DOI).