During normal cranial morphogenesis, most neurocranial sutures remain open till advanced adult stages, closing long after the brain is fully grown and developed. In some cases, however, sutures may close prematurely causing a pathological condition known as craniosynostosis which is associated with different neurocranial malformations depending on the suture involved. Apart from its aesthetic consequences, craniosynostosis is relevant from a biomedical perspective as malformations due to premature fusion of sutures may result in limited or abnormal brain growth and development, higher intra-cranial pressure, as well as respiratory and visual impairments. Moreover, a complete understanding on craniosynostotic processes is also relevant for evolutionary biologists, as species-specific cranial and cerebral morphologies may be somewhat influenced by changes on the particular tempo and sequence of suture closure (see for example the case of the australopithecine metopic suture).
From the above lines, it seems clear that an early and accurate diagnosis of the disease is essential for the correct management and prevention of possible clinical complications, as well as proper planning of surgical interventions. In this context, Carlos S. Mendoza and colleagues have recently published an automated pipeline that allows untrained observers to diagnose early craniosynostotic individuals with a 96% accuracy. In brief, their novel methodology is based on the automated computation of suture fusion indices and on the quantitative characterization of local deformation and curvature changes of bones and sutural areas. For the former task, the proposed pipeline incorporates a dedicated algorithm based on graph-cut techniques which is capable for automatically detecting and labeling both bones and sutures. After the identification of skeletal and sutural components, open and fused sutures are detected following a ‘contact degree’ criteria and fusion indices are computed as the number of voxels belonging to different bones that are in contact. On the other hand, normal neurocranial shape variation is modeled using a novel landmark-free methodology and principal component analysis (PCA). The resultant shape space is then used as a multi-atlas reference to compare and evaluate any potential patient that is to be diagnosed. Although this methodology may show some limitations, validation analyses presented by Mendoza and colleagues have demonstrated that by considering together fusion indices and shape descriptors it is possible to automatically discriminate between normal synostotic patients, as well as between different phenotypes of craniosynostosis. Moreover, this methodology not only enables the fast and accurate diagnosis of particular clinical cases (with all its associated medical advantages), but it also provides a methodological framework for the statistical analysis of large datasets, which in turn could provide important information regarding the structural and functional role of sutures during ontogeny and phylogeny.
José Manuel de la Cuétara