Abstracts

Rationale: Single Photon Emission Computed Tomography (SPECT) offers a unique advantage over other imaging modalities with the ability to image cerebral blood flow (CBF) during a seizure. It allows for non-invasive comparison between ictal and interictal brain activity. Previously, the Ictal-Interictal SPECT Subtraction Analysis by SPM (ISAS) algorithm was shown to be a successful tool in localizing focal epilepsy. However, the current ISAS algorithm has several disadvantages including challenging implementation with different SPM and MATLAB versions, less accurate registration, and a biased estimate of the population standard deviation. To overcome these problems, we implemented a modified version ISAS algorithm in BioImage Suite, an open source software package. We named this new variant of the ISAS algorithm Ictal-Interictal SPECT Subtraction Analysis by BioImage Suite (ISAB). Methods: The original ISAS method is biased as it uses a signed difference between the two sets of control SPECT images to estimate the normal scan variation in SPECT imaging. However, since scan order is not important for the control population, the results will differ if the scan order is changed. We have eliminated this bias by using a half normal distribution model to estimate the underlying distribution of normal scan variations. In addition, the original ISAS implementation uses a full general linear model (GLM) for detecting regions of abnormal CBF. With no loss of accuracy, we have reformulated this to use a simpler t-test implementation which compares the differences between an ictal and an interictal SPECT image for a patient to the differences between two SPECT images for each individual in a control population, resulting in a significant computational speed up. Multiple comparison correction and cluster-level statistic calculations were implemented by using Random Field Theory. Results: Initial results show that the new ISAB algorithm provides as good or possibly better localizing abilities as the original implementation of the ISAS algorithm. Image registration is improved, reducing extracranial skull artifacts. By eliminating the unnecessary GLM calculations, the amount of data storage and time needed for the statistical comparison has been reduced. Use of the unbiased standard deviation estimate in ISAB did not substantially change the results of seizure localization. The design of the GUI integrates into the Datatree Manager of BioImage Suite allowing for efficient organization of data, easy-to-use interface, and integration into the BrainLAB Vector Vision Cranial image-guided neurosurgery system. Conclusions: The reimplementation of the ISAS algorithm and the introduction of the ISAB algorithm in BioImage Suite provides a free/open source utility for subtraction-based SPECT analysis, which can be used on Windows/Mac/Linux operating systems with no commercial software prerequisites. Our goal is for ISAB is to make ictal-interictal SPECT analysis readily available in a statistically robust format for all epilepsy centers that perform ictal SPECT.