Surgical outcome in patients with nonlesional refractory partial epilepsies could be improved by a more precise definition of the epileptogenic zone (EZ). The value of interictal FDG-PET hypometabolism, voxel-based-morphometry (VBM) and diffusion tensor imaging (DTI) is still debated. We compared the sensitivity and specificity of these noninvasive techniques in localizing the EZ with stereo-electroencephalography (SEEG) results. Twenty patients with nonlesional partial epilepsy (13 temporal lobe epilepsy (TLE) and 7 extra-temporal (extra-TLE)) underwent structural MRI, DTI and FDG-PET. FDG-PET was analyzed visually (vPET) blinded and unblinded and by statistical parametric mapping (SPM) (sPET). Individual modifications of grey matter volume and mean diffusivity increase were compared to a control group with SPM. The best sensitivity was provided by vPET unblinded (75%) and the best specificity (60%) by DTI. The sensitivity of vPET blinded (55%) was lower and those of sPET (40%) and VBM (35%) were still lower. In TLE, vPET analyzed either blinded or unblinded, performed the best and additional use of the other tools improved slightly the sensitivity. For extra-TLE, combining vPET and DTI results increased the number of pertinent abnormalities detected especially for circumscribed changes in frontal lobe epilepsy (FLE). Combining vPET and DTI was the more efficient strategy for extra-TLE, allowing the detection of pertinent abnormalities in FLE when FDG-PET alone was not contributive. Combining sPET or VBM with vPET was less useful.