Deep Hybrid Learning Method for Classification of Fetal Brain Abnormalities

In recent years, lot of work has been carried out to develop a computer automated system to identify brain disorders. In the study and research of fetal brain disorders MRI images plays vital role. From the study of several literatures it is observed that existing machine learning techniques for the classification of fetal brain MRI are complex, time consuming and facing the problem of over-fitting. In the proposed system Deep Hybrid Learning (DHL) method is used for classification of fetal brain abnormality. In this work, the fusion of Deep Learning technique with the conventional machine learning method has been carried out in order to obtain the good classification results. The aim of this research study is to make more acceptable results in the classification of fetal brain abnormality using MRI images. The classification layer of Deep Neural Network (DNN) architecture is replaced by Random Forest (RF) machine learning classifier. The experimental results obtained from DNN+RF model are compared with the results of simple DNN and DNN+SVM framework. It shows that the proposed system achieves the good classification result. The DNN+RF has an Area Under Curve (AUC) of 94% and 87% for training and validation respectively which is better than the state-of-arts method. The paper is concluded with challenges and possible future directions.