# Optimising Gallium-68 (⁶⁸Ga) DOTATATE PET/CT Reconstruction in Neuroendocrine Tumours: A Paired Comparison of Penalised-Likelihood (BSREM/Q.Clear) and Ordered Subset Expectation Maximisation Algorithms

## TL;DR

This study compares two image reconstruction methods for Ga-68 DOTATATE PET scans in neuroendocrine tumors, finding that the Bayesian Penalised Likelihood method improves image quality and signal metrics.

## Contribution

The study evaluates the Bayesian Penalised Likelihood algorithm for Ga-68 DOTATATE PET scans in neuroendocrine tumors, identifying optimal parameters for improved image quality and quantitative metrics.

## Key findings

- BPL reconstruction increased lesion SUVmax and SNR compared to VPFX across all β values.
- Qualitative image quality was best at β = 800 with lower artefact scores.
- BPL is a viable alternative for Ga-68 DOTATATE PET reconstruction in neuroendocrine tumors.

## Abstract

Objective

The aim was to determine whether “Q.Clear” (GE Healthcare, Bayesian Penalised Likelihood (BPL) reconstruction algorithm) of Gallium-68 (⁶⁸Ga) DOTATATE PET scans at different penalisation factors (β) could improve qualitative and quantitative image parameters compared with the standard Ordered Subsets Expectation Maximisation (OSEM), VPFX reconstruction.

Methods

Twenty-five PET/CT scans performed 60 minutes after injection of 110-224 MBq (activity 153 MBq/kg) of ⁶⁸Ga-DOTATATE on a GE Discovery 710 PET/CT scanner were reconstructed using VPFX (2 iterations, 24 subsets) and Q.Clear with β values ranging from 200-1200.

A representative neuroendocrine tumour (NET) lesion and three reference regions (liver, spleen, and L3 bone marrow) were measured for standardised uptake values (SUVₘₐₓ/mean/peak/SD), signal-to-noise ratio (SNR = SUVₘₐₓ/liver SUVSD), and signal-to-background ratio (SBR = SUVₘₐₓ/liver SUVmean). A blinded qualitative assessment by a PET specialist scored image quality on a 5-point scale and evaluated the presence and severity of artefacts.

Results

BPL lesion SUVₘₐₓ and SNR were greater than VPFX for all β values (p < 0.05). Although BPL lesion SBR values were higher than VPFX, no β reached statistical significance. Similar patterns were observed for reference organ comparisons. Qualitative analysis showed a preference for β = 800, which yielded the best image quality with lower artefact scores.

Conclusion

BPL reconstruction of ⁶⁸Ga-DOTATATE PET data in patients with NETs improves SNR in tumour lesions and normal organs and increases SUVₘₐₓ in tumours. Combining these results with the preferred image quality at β = 800, BPL reconstruction can be considered a viable alternative for future reconstruction methods when assessing NETs.