Computer-aided automatic transfer learning based approach for analysing the effect of high-frequency EMF radiation on brain

An electromagnetic field radiations (EMF) emanating from the cell phones affect the brain and other organs in living organisms. Therefore, the objective of the present study is to examine whether the EMF radiations affect the brain cells or not, using the transfer learning-based methodology. The observations made in the present study are based on our experiment with the Drosophila melanogaster. The microscopic brain-images of drosophila, (should be read as micro-images) exposed to and not exposed to the high-frequency EMF radiations were acquired for the analysis purposed. The comprehensive set of features were extracted from the micro-images in EMF-exposed and Non-exposed class drosophila using the pre-trained convolution neural networks (CNNs). Further, the support Vector Machine (SVM) has been used to find an optimal hyperplane in higher dimensional feature space which separates the feature representations of the micro-images from both the classes. The % accuracy of SVM in classifying the features extracted from the micro-images from both the classes using the pre-trained VGG19 network is 87.3% for the 5-fold cross-validation. The discrimination in feature sets extracted from the micro-images signifies that the EMF radiations have affected the drosophila brain cells. Experimental results reveal that the prolonged exposure to EMF radiations might have affected the drosophila brain which approves the assumed hypothesis.