MALARIA DETECTION IN HUMAN CELLS USING DEEP LEARNING

Abstract

This master’s report presents an automatic identification of Malaria infected cells in human blood samples using deep learning methods. We used a dataset of 27,558 images divided equally into infected and uninfected cells. Various machine learning and deep learning techniques used to solve this task. However, practical performance has not proven sufficient yet. In this respect, we built a system that can discriminate the infected cells from the uninfected ones. We proposed three different deep learning architectures, convolutional neural network (CNN), AutoEncoder and transfer learning using pretrained model (VGG16). The best result went to the CNN model which showed superiority over the AutoEncoder and transfer learning model (VGG16) in all performance indicators such as sensitivity (99.78%), specificity (99.95%) precision (99.87%) and F1 score (99,87%).