Purpose: We aim at studying intrinsic structures of the intracranial subarachnoid space.

Material and methods: Magnetic resonance images were obtained using the SPACE sequence, and the segmentation of the superior intracranial subarachnoid space was performed using geometrical features and a topological assumption of the shapes. Given such segmentation results, we present a method based on a geodesic propagation technique, which allows us to make an intrinsic profile of the space. Intrinsic profiles are then analyzed qualitatively and quantitatively, in particular for classification into healthy and pathological cases based on their intrinsic bilateral asymmetry and histogram moments.

Results: The proposed method was applied to a clinical dataset of 15 subjects, of which 7 were healthy volunteers and 8 were hydrocephalus patients. The intracranial cerebrospinal fluid is not (intrinsically) bilaterally asymmetric for healthy volunteers, while hydrocephalus would cause asymmetry. We also observed that the results of a two-class classification (healthy or not) based on histogram moments were suitable; sensitivity, specificity and precision are all 100%.

Conclusions: The effectiveness of the proposed method of intrinsic profiling analyses is shown by preliminary experiments on healthy adults and hydrocephalus patients.