2 million people die every year due to Malaria disease. Moreover, it is estimated that there are 300-500 million new cases annually. It means massive data load that hospitals and laboratories have to manage and process, while some regions that are suffering from the disease are often lacking well-trained personnel that can perform high-quality microscopy examination and that’s due to high costs to train such experts. Apart from that, the whole examination process can be also very time-consuming and error-prone and that’s our problem that we want to solve.
The model that we’re creating and improving is basing on Malaria Dataset which is a repository of segmented cells from the thin blood smear slide images from Malaria Screener research activity. Malaria Screener is an already existent technology- an application with different modules to acquire, manage and visualise thin and think blood smear images of P. falciparum parasites.
Our project targets to develop a classification model with binary problem to accurately detect malaria disease present in images of thin blood smears. To do so we have used a total of 27,558 cell images with equal instances of parasitized and uninfected cells. We created a Convolutional Neural Network using Tensorflow and Keras which are able to classify whether the cells are infected or not.
The goal of using technology and machine learning here is to make the microscopy examination process faster, more consistent, and less dependent on human expertise thus helping world detecting and fighting malaria in urgent situations.
https://lhncbc.nlm.nih.gov/LHC-downloads/downloads.html#malaria-datasets --> click on the link and download cell_images.zip from there to access segmented cells