Evaluation of ictal brain SPECT using statistical parametric mapping in temporal lobe epilepsy

An automated voxel-based analysis of brain images using statistical parametric mapping (SPM) is accepted as a standard approach in the analysis of activation studies in positron emission tomography and functional magnetic resonance imaging. This study aimed to investigate whether or not SPM would increase the diagnostic yield of ictal brain single-photon emission tomography (SPET) in temporal lobe epilepsy (TLE). Twenty-one patients (age 27.14 +/- 5.79 years) with temporal lobe epilepsy (right in 8, left in 13) who had a successful seizure outcome after surgery and nine normal subjects were included in the study. The data of ictal and interictal brain SPET of the patients and baseline SPET of the normal control group were analysed using SPM96 software. The t statistic SPM¿t¿ was transformed to SPM¿Z¿ with various thresholds of P<0.05, 0.005 and 0.001, and corrected extent threshold P value of 0.05. The SPM data were compared with the conventional ictal and interictal subtraction method. On group comparison, ictal SPET showed increased uptake within the epileptogenic mesial temporal lobe. On single case analysis, ictal SPET images correctly lateralized the epileptogenic temporal lobe in 18 cases, falsely lateralized it in one and failed to lateralize it in two as compared with the mean image of the normal group at a significance level of P<0.05. Comparing the individual ictal images with the corresponding interictal group, 15 patients were correctly lateralized, one was falsely lateralized and four were not lateralized. At significance levels of P<0.005 and P<0.001, correct lateralization of the epileptogenic temporal lobe was achieved in 15 and 13 patients, respectively, as compared with the normal group. On the other hand, when comparison was made with the corresponding interictal group, only 7 out of 21 patients were correctly lateralized at the threshold of P<0.005 and five at P<0.001. The result of the subtraction method was close to the single case analysis on SPM at P<0.05. However, at higher thresholds (P