Evaluation of an anatomical based MAP reconstruction algorithm for PET in epilepsy

We studied the performance of an anatomical based maximum-a-posteriori reconstruction algorithm (A-MAP) for the detection of hypo-metabolic regions in positron emission tomography (PET) of the brain of epilepsy patients. Between seizures, 2-[¹⁸F]fluoro-2-deoxy-D-glucose PET shows a decreased glucose metabolism in gray matter (GM) associated with the epileptogenic region. However, detection of these regions is limited by noise in the measurement and the relatively small thickness of GM compared to the spatial resolution of PET. We hypothesized that incorporating anatomical information, derived from magnetic resonance imaging data, and pathophysiological knowledge in the reconstruction process improves the detection of hypo-metabolic regions. Monte-Carlo based brain software phantom experiments were used to examine the performance of A-MAP. The influence of small misregistration errors of the anatomical information and weight of the a priori information in GM were studied. A-MAP showed improved results for signal-to-noise ratio, bias and variance. A human observer study was performed, showing improved detection of hypo-metabolic regions using A-MAP compared to maximum-likelihood (ML) reconstruction. Finally, A-MAP was applied to clinical PET data of an epilepsy patient. We can conclude that the use of anatomical and pathophysiological information during the reconstruction process is promising for the detection of subtle hypo-metabolic regions in the brain of patients with epilepsy.