SPECT Reconstruction With a Trained RegularizerUsing CT-Side Information: Application to177LuSPECT Imaging

초록

Improving low-count SPECT can shorten scans andsupport pre-therapy theranostic imaging for dosimetry-basedtreatment planning, especially with radionuclides like (177)Luknown for low photon yields. Conventional methods often under-perform in low-count settings, highlighting the need for trainedregularization in model-based image reconstruction. This articleintroduces a trained regularizer for SPECT reconstruction thatleverages segmentation based on CT imaging. The regularizerincorporates CT-side information via a segmentation mask froma pre-trained network (nnUNet). In this proof-of-concept study,we used patient studies with( 177)Lu DOTATATE to train andtested with phantom and patient datasets, simulating pre-therapyimaging conditions. Our results show that the proposed methodoutperforms both standard unregularized EM algorithms and con-ventional regularization with CT-side information. Specifically, ourmethod achieved marked improvements in activity quantification,noise reduction, and root mean square error. The enhanced low-count SPECT approach has promising implications for theranosticimaging, post-therapy imaging, whole body SPECT, and reducingSPECT acquisition times.

키워드