Robust anatomy recognition for automated MR neuro scan planning
Netsch T, Bergmans R, Gieseke J, Hoogenraad F, van Muiswinkel A, Visser F, Springorum R, Young S, Bystrov D
A novel method for automated planning of scan geometries for MR scans of the brain is presented. The approach is based on the robust identification of a set of anatomical landmarks in a 3D survey image. A clinical trial was carried out over a six-month period, in which the reliability of the system was tested. Results are presented to demonstrate the accuracy of estimation of the mid-sagittal plane and landmark estimation. The approach was shown to be very robust, successfully planning geometries for all patients, including in the presence of large tumors, significant metal-induced artefacts and in patients with extensive brain resections. Over 230 clinical patients were scanned in the trial, and in each case the system provided a suitable initialization of the planning. There were no outliers observed.