PT - JOURNAL ARTICLE AU - Alexander Hans Vija AU - Michal Cachovan TI - <strong>Multi-modal Reconstruction in Brain Perfusion SPECT</strong> DP - 2019 May 01 TA - Journal of Nuclear Medicine PG - 1362--1362 VI - 60 IP - supplement 1 4099 - http://jnm.snmjournals.org/content/60/supplement_1/1362.short 4100 - http://jnm.snmjournals.org/content/60/supplement_1/1362.full SO - J Nucl Med2019 May 01; 60 AB - 1362Objectives: Extra-modal information from structural modalities such as CT and MR can help improve resolution and overall image quality of SPECT reconstructions. We developed a prototype reconstruction for SPECT neurology that incorporates MR information into SPECT reconstruction. The aim of this study is to estimate the resolution improvement compared to legacy reconstructions in a small set of brain perfusion studies. Methods: Ten patients and their datasets were included in this study. Each acquisition dataset package consists of list mode based SPECT projection data, a CT and an MR T1-weighted dataset of the head. MR was segmented into cerebro-spinal fluid, grey and white matter and remaining tissue, allowing for a mixture of tissues especially at tissue borders. SPECT projection data were reconstructed with 1.95mm voxel size using product xSPECT Quant reconstruction (xQ) and a prototype xSPECT Neurology multi-modal reconstruction (xN) with 40 conjugate gradient updates, including attenuation/scatter correction and quantitative calibration, with 10 mm post-smoothing. Using a matched filter approach, we minimized the difference between xQ and xN images, where the xN images were smoothed using a 3D-Gaussian filter with varying FWHM between 0 and 20 mm. The FWHM which minimized the absolute marginal distance is a global matched filter that indicates the resolution improvement between xQ and xN. Results: For the ten datasets enrolled in this study the resulting matched filter was 8.5 +/- 0.3 mm. Absolute difference between the xQ image and the matched filter convolved xN image averaged to 11.3% +/- 2.1%. The average difference of the images was -3.2% +/- 2.3%. The R2 for the resulting matched filter convolved images averaged to 0.78 +/- 0.05. Conclusions: Multi-modal reconstruction in brain perfusion imaging is feasible using clinical data and improves image resolution. If the MR assisted xN images are post smoothed by a 3D Gaussian filter with 8.5mm FWHM, then the difference to the xQ image is minimized on average. This is an important step to assure that a clinical read based off this new reconstruction can be related back to clinically used images. Next step is to assess the clinical impact of the MR assisted xN images in best resolution as compared to xN with the post filter and xQ in diagnostics of neurodegenerative diseases as well as epilepsy.