About This Project

We are developing a frugal diagnostic for urogenital schistosomiasis, a neglected tropical disease that is responsible for over 200,000 deaths annually. Building off of the Planktoscope technology, we are working on a flow-through system that images patient samples and can identify infection via machine learning. This work is important because compared to the current method of diagnosis, our method seeks to improve the sensitivity and cost of diagnosis and allow for real-time detection.

Ask the Scientists

What is the context of this research?

Schistosomiasis is the second most prevalent parasitic disease in the world with high rates of infection in Sub-Saharan Africa. Currently, the main method of diagnosis is by urine filtration microscopy, which is a time-intensive and labor-intensive technique that has a low sensitivity for infection. This technique has a sensitivity of about 54% as many of the eggs may not be caught by the filter and can also be missed when counting by hand. With our system, we are trying to eliminate the need for filtration and counting eggs by utilizing a system that pumps the sample while simultaneously taking a video capture of the sample. This, paired with a machine learning model that can segment and classify eggs, would be able to sensitive and rapid diagnosis.

What is the significance of this project?

A system that can measure schistosomiasis sensitively, rapidly, and frugally could help people in rural and low-income areas get access to diagnosis. Ultimately, better diagnosis and data on schistosomiasis could help direct more efforts and funding towards treatment and prevention efforts.

What are the goals of the project?

Our overall goal for this project is to develop a system that is sensitive, rapid, frugal, and easily used. With this in mind, we will develop a robust data set of schistosomiasis haematobium eggs to be able to create a machine learning model to segment and classify levels of infection. We also will optimize parameters for our flow-through system and test out different designs. From there, we will work on the model to ensure that it is well-defined even for light infections of schistosomiasis in a variety of conditions and that it can segment and classify in real-time.