In pediatric ds005083, independent SEEG CT contact-centre error was 0.246 mm 3D RMS, with 1.31-fold greater error along the shaft

We measured the independent localization error of SEEG contact centres in OpenNeuro ds005083 v1.0.0 (58 subjects, 615 shafts, native CT). The method uses the constant contact pitch. For each inner contact, the second difference C(i+1) - 2C(i) + C(i-1) is almost 0 on a straight shaft. Its spread thus gives the error of the centres. The median 3D RMS error per contact was 0.246 mm (95% CI 0.234 to 0.258 over subjects). The axial SD was 0.159 mm and the lateral SD was 0.128 mm per axis. The ratio was 1.31 (CI 1.18 to 1.38). It was larger than 1 in 50 of 58 subjects (sign-flip p = 5e-5). Deeper contacts had slightly larger error (median Spearman rho -0.085, p = 5e-5). Of 6649 adjacent spacings, 0.93% were more than 1 mm from the shaft median spacing. The error of a contact-to-contact distance is approximately 0.20 mm SD. Thus 3.2% of pairs go into the wrong 5 mm distance bin and 8.1% into the wrong 2 mm bin. The method does not measure shaft-wide shifts, CT-to-MRI registration error or brain shift. Thus the value is a lower bound on the total error. The estimator and tests were set in PLAN.md before the analysis was run. The package contains code, an MD5 manifest of the inputs, the results and a figure.

stereo-electroencephalography · SEEG · contact localization · CT · localization error · distance measurement · pediatric neurosurgery