Background: Manual segmentations of the whole intracranial vault in high-resolution magnetic resonance images are\noften regarded as very time-consuming. Therefore it is common to only segment a few linearly spaced intracranial areas\nto estimate the whole volume. The purpose of the present study was to evaluate how the validity of intracranial\nvolume estimates is affected by the chosen interpolation method, orientation of the intracranial areas and the linear\nspacing between them.\nMethods: Intracranial volumes were manually segmented on 62 participants from the Gothenburg MCI study using\n1.5 T, T1-weighted magnetic resonance images. Estimates of the intracranial volumes were then derived using\nsubsamples of linearly spaced coronal, sagittal or transversal intracranial areas from the same volumes. The subsamples\nof intracranial areas were interpolated into volume estimates by three different interpolation methods. The linear\nspacing between the intracranial areas ranged from 2 to 50 mm and the validity of the estimates was determined by\ncomparison with the entire intracranial volumes.\nResults: A progressive decrease in intra-class correlation and an increase in percentage error could be seen with\nincreased linear spacing between intracranial areas. With small linear spacing (?15 mm), orientation of the intracranial\nareas and interpolation method had negligible effects on the validity. With larger linear spacing, the best validity was\nachieved using cubic spline interpolation with either coronal or sagittal intracranial areas. Even at a linear spacing of\n50 mm, cubic spline interpolation on either coronal or sagittal intracranial areas had a mean absolute agreement\nintra-class correlation with the entire intracranial volumes above 0.97.\nConclusion: Cubic spline interpolation in combination with linearly spaced sagittal or coronal intracranial areas overall\nresulted in the most valid and robust estimates of intracranial volume. Using this method, valid ICV estimates could be\nobtained in less than five minutes per patient.
Loading....