Introduction: Annually, hydrocephalus affects nearly 7 children per 10,000 live births around the world. It significantly impairs the quality of life of such children and is associated with increased morbidity and mortality The high cost of treatment and post-intervention complications add to the burden of disease. Deployment of machine learning (ML) models in actual clinical settings have led to improved outcomes.
Objective: The aim of this systematic review is to analyze the utility as well as acknowledge the achievements of AI/ML in HCP decision making.
Methodology: PubMed and Cochrane databases were used to perform a systematic search with proper terminology to include all the relevant articles up to May 2021.
Results: Fifteen studies that described the use of ML models in the diagnosis, treatment, and prognostication of pediatric hydrocephalus were identified. The median accuracy of prediction by the ML model in various tasks listed above was found to be 0.88. ML models were most commonly employed for ventricular segmentation for diagnosis of hydrocephalus. The most frequently used model was neural networks. ML models attained faster processing speeds than their manual and non-ML-based automated counterparts.
Conclusion: This study attempts to evaluate the important advances and applications of ML in pediatric hydrocephalus. These methods may be better suited for clinical use than manual methods alone due to faster automated processing and near-human accuracy. Future studies should evaluate whether the use of these models is feasible in the future for patient care and management in field settings.
Keywords: Artificial intelligence; hydrocephalus; machine learning; neural networks; pediatric.