Developing a machine learning model for malaria diagnosis in rural areas
Date
2018-04
item.page.datecreated
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Medical diagnosis of diseases like Malaria and tuberculosis still use microscopy as a
standard, but this procedure is usually very tiring for pathologists and health workers as it
imposes much stress on their vision. Due to the fatigue that health workers get from this process,
they might end up misdiagnosing a case. In most Rural areas of Cameroon and Ghana, there are
no qualified personnel to do these diagnoses. Moreover, according to the World Bank, malaria
still kills millions of people every year in Sub-Saharan Africa. To solve this problem, we used a
machine learning approach; transfer learning to retrain an already existing model to perform
binary classification on malaria blood smear images. The pretrained model was already
optimized for devices with low memory, therefore this project’s model can work on low memory
devices with no network connectivity. This project also explored Generative Adversarial
networks as an alternative way of training a classifier for scenarios with data scarcity. This
project shows how a model trained on a different task can be retrained to solve a similar task and
shows a technique for developing a classifier in scenarios of data scarcity.
Description
Applied project submitted to the Department of Computer Science, Ashesi University, in partial fulfillment of Bachelor of Science degree in Computer Science, April 2018
item.page.type
Applied project
item.page.format
Keywords
malaria diagnosis, blood smear images, machine learning algorithm, computer vision software, mobile application