Skip to main content
 

Computational Image Analysis and Deep Learning for Malaria Parasite Detection and Quantification

Wednesday, September 12, 2018 — Poster Session II

3:30 p.m. – 5:00 p.m.
FAES Terrace
NLM
MICROBIO-11

Authors

  • M Poostchi
  • S Jaeger
  • G Thoma

Abstract

Malaria remains a major burden on global health, with roughly 216 million cases worldwide and more than 440,000 deaths in 2016. It is caused by parasites that are transmitted through the bites of infected female Anopheles mosquitoes, which infect the red blood cells. Most deaths occur among children in Africa, where malaria is a leading cause of childhood neuro-disability. The gold standard for malaria diagnosis in the field is light microscopy of blood smears, where an expert microscopist visually inspects blood smears for parasites. Accurate parasite counts are essential not only for malaria diagnosis. They are also important for measuring drug-effectiveness and classifying disease severity. However, manual counting of parasites in hundreds of millions of blood smears worldwide every year is a laborious, costly and error-prone process that depends heavily on the experience and the skill of the microscopist. We show that modern information technology and machine learning techniques can play key roles in fighting the disease and improving diagnosis. In particular, we develop an automatic system to identify and quantify malaria parasites in real-time on blood smears. The malaria parasite detection system consists of three processing steps: (i) image acquisition of blood smear images using a standard light microscope with an attached smartphone camera, (ii) detection and segmentation of blood cells, and (iii) cell classification. We present a microscopy image analysis framework for multi-cell detection and malaria parasite stage identification and quantification using deep neural network.

Category: Microbiology and Infectious Diseases