Diagnosis of Malaria Infected Blood Cell Digital Images using Deep Convolutional Neural Networks

Abstract

Automated medical diagnosis is an important topic, especially in detection andclassification of diseases. Malaria is one of the most widespread diseases, with morethan 200 million cases, according to the 2016 WHO report. Malaria is usuallydiagnosed using thin and thick blood smears under a microscope. However, properdiagnosis is difficult, especially in poor countries where the disease is mostwidespread. Therefore, automatic diagnostics helps in identifying the diseasethrough images of red blood cells, with the use of machine learning techniques anddigital image processing. This paper presents an accurate model using a DeepConvolutional Neural Network build from scratch. The paper also proposed threeCNN models each one trained on the Malaria RBC dataset with differentarchitectures for handling the classification tasks. Furthermore, disadvantage of thetraditional method of using transfer learning, and how to control model complexityto achieve better performance was discussed. The dropout regularization techniquewas used to avoid overfitting problems and minimize validation loss. Applying DataAugmentation technique to avoid the problem of small data in training of proposedmodels, which is a very common problem in medical dataset. Finally, removingnoise in Malaria images using a Median blur filter, and studying how effects of thaton training CNN models. According to the classification results, the proposed modelachieved better classification results at accuracy 99.22 on the original Malaria RBCsdataset, and it has the best performance comparing with related work.