• Michael Nelles, Juergen Gieseke, Horst Urbach, Renate Semrau, Daniel Bystrov, Hans H Schild, and Winfried A Willinek.
• Department of Radiology, University of Bonn, Bonn, Germany. michael.nelles@ukb.uni-bonn.de
• J Magn Reson Imaging. 2009 Sep 1; 30 (3): 672-7.

PurposeTo evaluate the feasibility of automatic planning and scanning of brain MR imaging (MRI) protocols on a clinical 3 Tesla system in tumor patients before and after neurosurgical intervention.Materials And MethodsTwenty-nine patients with intra-axial lesions were examined with automated planscan software pre- and postoperatively. MR section geometries were determined using intensity-based three-dimensional registration and an extraction of landmarks. The technique involved an active shape model to match the boundaries of anatomical structures and typical shape variations. Insufficient geometries were corrected manually by a trained operator.ResultsIn 29/29 of the preoperative and 47/58 MRI sessions in total, no manual interaction was necessary. Predominantly minor corrections were necessary in 11/29 postoperative sessions, with critical corrections (> or = 3-mm offcenter change or > or = 5 degrees in alignment of the stacks) in 3/58 sessions. Mean offcenter correction was 1.41 mm (range, 0-7.33 mm), mean angle change toward the midline or commissural line was 1.43 degrees (range, 0-8.05 degrees ).ConclusionAutomatic planning and scanning before and after brain surgery yields robust results in most of the patients with substantial shape deviations. The dimensions of necessary geometry corrections are predominantly small. These results are promising to minimize interscan variability in longitudinal studies.

31 keywords...

hide…