Reader dependent and independent methods of visual and quantitative analysis in detection of epileptic focus: Visual and quantitative approaches to brain 18F-FDG PETt using statistical parametric mapping (SPM)

Interictal brain 18F-FDG-PET is used in medically refractory epilepsy patients in preparation for surgical treatment in order to localize the epileptogenic focus. However, in the interictal brain 18F-FDG-PET images hypometabolism may also be observed in cortical regions where the seizure activity propagates and this makes the determination of epileptogenic focus difficult. In order to overcome this difficulty objective quantitative analytic methods are applied that may contribute to the visual assessment of brain 18F-FDG-PET images in epilepsy patients. Statistical parametric mapping is one of these quantitative analytic methods.In this study we aimed to evaluate differences between visual assessment and quantitative analysis in the detection abnormality (lateralization in epilepsy) with the application of reader dependent and independent methods and to see how these differences are shaped depending on the experience of readers.Interictal brain 18F-FDG-PET data of 54 temporal lobe epilepsy patients who were operated and classified as Engel I on follow-up were retrospectively analyzed. In addition brain 18F-FDG-PET images of 16 patients who had undergone whole body 18F-FDG-PET examinations due to a solitary pulmonary nodule or the primary staging of a malignant disease and who did not have additional neurological disease or received any prior therapy was used as the control data. Presurgical interictal brain 18F-FDG-PET images were evaluated by three physicians with different levels of expertise on two occasions: once blind to SPM results and secondarily together with patients? SPM results. The two statistical thresholds used in SPM analyses were p<0,001 and p<0,01.Readers? assessments independent of SPM results were correct in 67%, 91% and 94% of patients, in the ascending order of expertise. When the SPM results were available during visual assessment, these ratios increased to 96%, 100% and 100%, respectively. Readers? mean levels of confidence in the SPM-independent assessment were 1.24±0.13, 1.59±0.09 and 1,80±0,07. These mean values increased to 1.89±0.06, 1.89±0.04 and 1.96±0.03, respectively.The sensitivity and specificity of SPM results at the statistical threshold level of p<0,001 were 78% and 96%. At the statistical threshold level of p<0,01 the sensitivity and specificity were found to be 94% and 74%, respectively. No significant correlations were found between the seizure duration and bilateral and extratemporal hypometabolism detected with SPM analysis at the statistical threshold level of p<0,01. However, a weak correlation was observed between the seizure duration and hypometabolic cluster extension. When both right and left temporal lobe epilepsy patients? data were compared with the normal data in separate clusters using SPM at the statistical threshold level of p<0,01, significant hypometabolism was found in the ipsilateral temporal, frontal lobes and the ipsilateral thalamic nucleus.As a conclusion, SPM helped the less experienced reader correctly lateralize the epileptogenic focus. In addition, SPM caused increases in the confidence levels in lateralizing the epiloptogenic focus of all readers with different levels of expertise.Regarding the weaknesses and strengths of SPM analyses at different statistical threshold levels in both evaluation of a single patient?s data and cluster analysis, the necessity to find optimum statistical threshold levels in accordance with the aim of the investigator comes into prominence.Key words: SPM, temporal lobe epilepsy, brain 18F-FDG PET