Deep learning neural networks in malaria diagnosis

Abstract

Malaria is a serious disease mostly spread in tropical and subtropical areas that causes 438.000 deaths per year. Current malaria diagnosis relies primarily on microscopic examination of stained blood films. This method is time consuming and prone to human error, even in experienced hands. Thus, there is a need for the development of an automatic technique that is able to detect malaria in a sensitive and unsupervised manner. Deep learning networks are a novel field that promises to have a key role in this automatic detection. In this thesis, we propose a system that collects much of the research conducted about this issue and that proposes new schemes to enhance the performance. In particular, a solution based on convolutional neural networks has shown a clear improvement of the results in the detection of malaria.