Introduction: Cranisynostosis is a rare, congenital anomaly, characterised by premature fusion of the calvarial sutures. Cranisynostosis is often associated with increased intracranial pressure, abnormal head shapes and developmental delays. Despite surgery being the first-line intervention, there is significant debate surrounding optimal operative timing, surgical technique, and post-operative care. This, combined with the condition's complex aetiology, often results in unpredictable functional and morphological outcomes. The use of digital technologies may help improve outcomes of craniosynostosis correction.
Methods: A comprehensive computational pipeline to evaluate and optimise spring-assisted cranioplasty configurations was studied in sagittal synostosis and posterior vault expansion. Initially, virtual populations of paediatric skulls were generated from a parametric skull model and a statistical shape model to represent anatomical variations, cranial size, morphology, and bone stiffness. Finite element analysis simulations were conducted across several surgical configurations to assess the biomechanical impact of varying surgical parameters alongside calvarial bone properties. Data analysis, utilising sensitivity indices and partial regression, evaluated the isolated and combined interrelations between these variables and the resulting surgical outcomes. Finally, explainable machine learning algorithms were trained on the simulated datasets to accurately predict postoperative cranial indices and automate optimal, patient-specific surgical planning.
Results: Results demonstrated that calvarial properties, specifically bone thickness and elastic modulus, were the most influential parameters on postoperative skull expansion in sagittal synostosis. The XGBoost machine learning algorithm accurately predicted postoperative intracranial volume and cephalic indices with 97% accuracy. The multi-objective optimisation successfully identified patient-specific parameter combinations that maximised volume while maintaining aesthetic feature angles below a threshold in posterior vault expansion.
Conclusions: Digital technologies such as Digital Twins and Machine Learning may help improve outcomes of craniosynostosis correction by providing personalised care and improving intraoperative precision. Additionally, virtual populations may allow researchers to safely conduct simulated clinical trials to test and optimise new surgical techniques.