On the optimal z-score threshold for SISCOM analysis to localize the ictal onset zone

Abstract Background In epilepsy patients, SISCOM or subtraction ictal single photon emission computed tomography co-registered to magnetic resonance imaging has become a routinely used, non-invasive technique to localize the ictal onset zone (IOZ). Thresholding of clusters with a predefined number of standard deviations from normality (z-score) is generally accepted to localize the IOZ. In this study, we aimed to assess the robustness of this parameter in a group of patients with well-characterized drug-resistant epilepsy in whom the exact location of the IOZ was known after successful epilepsy surgery. Eighty patients underwent preoperative SISCOM and were seizure free in a postoperative period of minimum 1 year. SISCOMs with z-threshold 2 and 1.5 were analyzed by two experienced readers separately, blinded from the clinical ground truth data. Their reported location of the IOZ was compared with the operative resection zone. Furthermore, confidence scores of the SISCOM IOZ were compared for the two thresholds. Results Visual reporting with a z-score threshold of 1.5 and 2 showed no statistically significant difference in localizing correspondence with the ground truth (70 vs. 72% respectively, p = 0.17). Interrater agreement was moderate (κ = 0.65) at the threshold of 1.5, but high (κ = 0.84) at a threshold of 2, where also reviewers were significantly more confident (p