Abstract ID: 247 Patient specific scatter reduction in SIRT gamma camera images

Clinical Bremsstrahlung imaging with gamma cameras or SPECT scanners, for example used in selective internal radiation therapy (SIRT) [1], suffers from low contrast due to a continuous spectrum and a high amount of scatter. Information about the scattering of the radionuclide within the patient's body can be obtained from Monte-Carlo simulations and subsequently being used to improve image quality.

An MAA-acquisition (CT+SPECT) of a HCC patient with a unifocal uptake in the right liver lobe is segmented (lesion) and loaded into the MC simulation framework GATE [2]. The voxelized lesion is used as Y90 source (1.5 GBq, 'fastY90' [3] is used yielding a speedup of 2.3×), the CT dataset is used for attenuation by converting HUs into corresponding materials. A mini gamma camera (Crystal Photonics, Germany) with a LEHR collimator and 4 × 4 cm² detector size is positioned close to liver on the patient's skin, pointing towards the lesion. A simulation (60 s acquisition time) is performed on a cluster with 512 cores (2.2-2.5 Ghz each). The total number of counts, the energy spectrum and the order of scattered particles within each geometric volume are obtained from the ROOT output. Particles that have not scattered at all are defined as primary events. Scattering is only calculated within the phantom.

The simulation took 172 min on the cluster with input data voxel size of 1 × 1 × 3.75 mm³. The number of emitted particles is 7.6 M with 600 k detected counts (∼8% ratio). 13.3% of the detected particles are primary events. Additionally, scatter of multiple orders has been observed (41%, 24% and 13% and 8% for the first four orders).

The energy-spectrum of the simulated 2D gamma camera image can be analyzed and used to correct the actual image acquired of that specific patient to achieve improved image quality and subsequently also dosimetry.