A dataset of blood slide images for AI-based diagnosis of malaria

Abstract

Malaria is a major public health challenge in sub-Saharan Africa. Timely and accurate diagnosis of malaria is vital to reduce the caseload and mortality rates associated with malaria. The use of microscopy in malaria screening is the gold standard recommended method by the World Health Organisation (WHO). In Uganda, utilization of microscopy is challenged by insufficient expertise to interpret the images accurately, affecting the efficiency, effectiveness and accuracy of malaria detection and diagnosis. We present a benchmark dataset of thick and thin blood smear images for automatic malaria screening in Uganda. Mobile Microscopy data was collected from Mulago Hospital, Department of Internal Medicine, Makerere University and Kiruddu National Referral Hospital in Uganda. The labelled image data can be used to build computational models implemented with convolution neural networks. The dataset has 3000 labelled thick blood smear images and 1000 labelled thin blood smear images. The datasets will support robust and accurate deep learning models for malaria diagnosis using thick and thin blood smear images with reasonable detection accuracies.