Cleft prediction before birth using deep neural network.

Area of Science:

• Medical Informatics
• Genetics
• Machine Learning

Background:

• Cleft lip and palate present significant challenges in developing countries due to high surgical costs and patient suffering.
• Early prediction of cleft conditions can facilitate timely interventions and improve outcomes.

Purpose of the Study:

• To develop and evaluate a machine learning model for predicting cleft lip and palate in embryos.
• To assess the feasibility of prenatal prediction to mitigate the impact of clefts.

Main Methods:

• Data collection from 1000 pregnant women in Lahore, Punjab, using a comprehensive questionnaire.
• Data preprocessing including cleaning, scaling, and feature selection.
• Comparative analysis of machine learning algorithms: random forest, k-nearest neighbor, decision tree, support vector machine, and multilayer perceptron.

Main Results:

• The multilayer perceptron (a deep neural network) demonstrated superior performance compared to other algorithms.
• The model achieved an accuracy of 92.6% on the test dataset for predicting cleft conditions.
• Key features influencing prediction were identified through feature selection methods.

Conclusions:

• Machine learning, particularly deep neural networks like multilayer perceptron, shows significant promise for prenatal prediction of cleft lip and palate.
• Accurate prediction can aid in early management strategies and potentially reduce the prevalence and impact of clefts.