Determination of Cranial Index in Sagittal Synostosis Using Artificial Intelligence Techniques

At least one main suture of the skull is closed prematurely in craniosynostosis, which may lead to different skull and face deformities and various types of child developmental delay. Increased posterior-anterior diameter of the skull is the main characteristic of sagittal craniosynostosis. The cranial index is the most important parameter for diagnosing and monitoring children with this deformity. This index is generally determined by direct measuring of infant’s head dimensions or using skull CT scan. Artificial intelligence-based techniques could identify the shape of the head from routine digital photos of a child and therefore, that may play a useful role in assisting physicians and parents with early diagnosis of skull anomaly, better intervention follow- up, and reducing medical system financial costs.

In this diagnostic study, by developing an executive scientific method, automatic measurement of the cranial index in sagittal craniosynostotic infants based on a simulated annealing algorithm was done (by using routine digital photos of their heads). Pre-operative photos of 59 patients operated in Mofid children hospital (Tehran, Iran); were processed, and the cranial index (normal value between 75% to 85%) was calculated with the proposed algorithm and compared with pediatric neurosurgeon measured values using Matlab software.

The proposed algorithm calculated the cranial index with acceptable accuracy. The simulated annealing algorithm determined the cranial index with the mean of 72.04 and the standard deviation of 5.21, which have minimal differences with surgeon-measured values with the mean of 72.21 and the standard deviation of 5.08. In the statistical investigation, using paired t-test, there was no statistically significant differences between these two methods (p-value = 0.52). The values measured by this method in 55 patients out of 59 studied patients (93.2%) are in the range of optimal values determined by the specialist (P ≤ 0.8).

It seems that the proposed method could determine the cranial index from routine digital patient’s head photos with acceptable accuracy. More development of this method may assist physicians and parents in early diagnosis of this anomaly, better monitoring of treatment results, and reducing medical financial costs.