18F-FDG PET studies in patients with extratemporal and temporal epilepsy: evaluation of an observer-independent analysis.
Image Processing, Computer-Assisted
Magnetic Resonance Imaging
Sensitivity and Specificity
Epilepsy, Temporal Lobe
The aim of this study was to evaluate an observer-independent analysis of 18F-fluorodeoxyglucose (FDG) PET studies in patients with temporal or extratemporal epilepsy.
Twenty-seven patients with temporal epilepsy and 22 patients with extratemporal epilepsy were included in the study. All patients with temporal epilepsy and 7 patients with extratemporal epilepsy underwent surgical treatment. In patients who showed significant postoperative improvement (temporal, n = 23; extratemporal, n = 6), the epileptogenic focus was assumed to be located in the area of surgical resection. In extratemporal epilepsy patients who did not undergo surgery, the focus localization was determined using a combination of semiology, ictal and interictal electroencephalography, [99mTc]ethyl cysteinate dimer SPECT, MRI and [11C]flumazenil PET. Visual analysis was performed by two experienced and two less experienced blinded observers using sagittal, axial and coronal images. In the automated analysis after anatomic standardization and generation of three-dimensional stereotactic surface projections (SSPs), a pixelwise comparison of 18F-FDG uptake with an age-matched reference database (n = 20) was performed, resulting in z score images. Pixels with the maximum deviation were detected, summarized and attached to one of 20 predefined surface regions of interest. For comparison with 18F-FDG PET and MR images, three-dimensional overlay images were generated.
In patients with temporal epilepsy, the sensitivity was comparable for visual and observer-independent analysis (three-dimensional SSP 86%, experienced observers 86%-90%, less experienced observers 77%-86%). In patients with extratemporal epilepsy, three-dimensional SSP showed a significantly higher sensitivity in detecting the epileptogenic focus (67%) than did visual analysis (experienced 33%-38%, each less experienced 19%). In temporal lobe epilepsy, there was moderate to good agreement between the localization found with three-dimensional SSP and the different observers. In patients with extratemporal epilepsy, there was a high interobserver variability and only a weak agreement between the localization found with three-dimensional SSP and the different observers. Although three-dimensional SSP detected multiple lesions more often than visual analysis, the determination of the highest deviation from the reference database allowed the identification of the epileptogenic focus with a higher accuracy than subjective criteria, especially in extratemporal epilepsy.
Three-dimensional SSP increases sensitivity and reduces observer variability of the analysis of 18F-FDG PET images in patients with extratemporal epilepsy and is, therefore, a useful tool in the evaluation of this patient group. The benefit of this analytical approach in patients with temporal epilepsy is less apparent.