Ictal SPECT is a highly sensitive method to localize the epileptogenic focus in refractory temporal lobe epilepsy in adults. In extratemporal epilepsy, sensitivity can be improved by subtracting interictal from ictal images and superimposing subtraction images on MRI. In children, such a procedure is potentially interesting because most epilepsies are extratemporal and ictal SPECT not yet routinely developed. The aim of this study was to test the feasibility of ictal SPECT with subtraction image processing in a pediatric population. Twenty-six children with refractory partial epilepsy and aged from 3 months to 18 years underwent ictal ECD-SPECT (20 mCi/1.73 m2) combined with video-EEG and interictal ECD-SPECT plus 3D-MRI two days later. Ictal-interictal subtraction images were computed by registering and normalizing the ictal to the interictal SPECT scans for each child. The ictal, interictal SPECT and subtraction images were registered to the children's MRI. Difference images were then superimposed to MRI for anatomical localization of the perfusion changes (overlay images). Looking for perfusion changes, overlay images allowed to detect at least one hyperperfused focus in 92 p. 100 of the 26 children compared to 73 p. 100 visually comparing ictal and interictal scans separately. Seizure onset was suspected on clinical and/or EEG and/or MRI in 19 children. Positive overlay images were concordant (n = 11) or larger (n = 7) than the suspected focus in 17/19 (90 p. 100), whereas they failed to show any abnormality in 1 child and were discordant with MRI in another one. In the 7 remaining patients, images showed cortical localization in 6 cases. Ictal SPECT is therefore faisable in very young children. Ictal-interictal subtraction SPECT images co-registered to MRI improves sensitivity compared to classical visual analysis. It seems therefore to be a helpful technique to localize the onset of seizure and to guide the intracranial recording in childhood epilepsy.