Detection of fetal brain abnormalities using data augmentation and convolutional neural network in internet of things

The all-embracing morphological changes in fetal brain development in the whole time pregnancy are visually seen in acquisition of Magnetic Resonance Imaging (MRI) techniques. In the way of exploring fetal brain development, Convolutional Neural Network (CNN) technique brings out automatic segmentation and classification. Monitoring fetal brain development is crucial to prevent brain abnormality. With the automation and resource optimization provided by IOT technology in healthcare applications, the national healthcare industries can improve the quality of care while lowering costs. The remarkable progresses in image recognition tasks are undergone via CNN to recognize complex patterns in image data. It faces the standstill challenges in observing the fetal brain development inclusive of 16–39 weeks gestation images and explores the optimized performance evaluation for quantitative assessment. The goal of this research is to improve access to brain development information by utilizing modern medical detection method as well as embedding IoT (internet of things). The brain malformations are considered as a problem in recent days, to overcome this problem the Detection of Fetal Brain Abnormalities (DFBA) using Data Augmentation is proposed in this paper and the IOT technology is helps to identify the detection in easy way. The proposed DFBA perform in three stages image pre-processing, model build and performance evaluation. The input dataset comprises 875 MRI scan images taken from the Harvard medical school with the training set as 80%, testing set as 10%, and validation set as 10%. Once the data is retrieved, the preprocessing techniques are carried out using Normalization and Reshaping for better progress and result. In sequence to pre-processing, Data augmentation is applied to increase the size of the dataset for CNN to be processed for classification. The efficiency of the proposed DFBA approach has been determined using the evaluation metrics such as Recall, F1-score, Accuracy, Precision, Confusion matrix and Support. The proposed method peculiarly tests set accuracy as 83% are obtained which is better than the existing classifiers