Non-invasive method for the prediction of micronutrient deficiency using sequential learning techniques

Abstract

ABSTRACT Study data from World Health Organization indicate that 2.8% of the disease around the world is caused by micronutrient insufficiency due to improper consumption of a balanced diet. The micronutrient deficiency is detected from the individual’s blood sample which is an invasive technique. Such techniques are quite in comfort for the persons and also painful. This proposed method involves the micronutrient deficiency identification using the non-invasive method of image processing. In this work, images of the anterior conjunctiva of eye, tongue and nail bed are used to detect the deficiency. Convolutional neural network models are used in the proposed deep learning model to extract the features of the parts of eye, nail and tongue. CNN-LSTM and CNN-GRU architectures are used to predict the deficiency in the images by extracting the features and perform the classification. A collection of 140 images were used as a dataset in this system. The real-time images are used to train the model, which is then validated with sample images for increased accuracy. The simulation of the CNN network with CNN-GRU has very low MSEs with 0.22%, results show that the proposed system achieved an accuracy of 99.6%, specificity of 99.7% and sensitivity of 99.4%.