Capstone Project Winner ParasiteID Aims to Facilitate Diagnoses in Developing World
While treatment of infections is a major hurdle in the developing world, the first step is diagnosis, which usually requires highly trained microscopists and expensive microscopes. ParasiteID aims to remove that barrier and facilitate effective diagnosis using machine learning methods on mobile cell phone microscopes that can be used by anyone.
ParasiteID is a MIDS Capstone Project from Fall 2018 and winner of the Hal R. Varian Award. MIDS graduates Cameron Bell, Vicki Foss, Kiersten Henderson, and Nathaniel Schub created the project and continue to develop it for its intended purpose of helping facilitate diagnoses of parasitic worm infections.
Tell us more about your project.
Vicki: ParasiteID is a tool designed to facilitate the diagnosis of parasitic worm infection in the developing world. The ParasiteID web application was designed to be coupled with inexpensive and highly mobile cell phone microscopes that can be operated by almost anyone, so that a user can easily upload a microscopy image of a patient stool sample and receive a prediction in real time about the parasite egg(s) present in the sample, based on an object localization model and an ensemble classification model. Our highly sensitive and specific classifier was created by assembling the largest parasitic worm egg image dataset of its kind to date. Our goal with ParasiteID is to enable more people suffering from parasitic infection in the developing world to be diagnosed and treated faster. ParasiteID reduces the bottleneck to patient treatment by removing the need for highly trained microscopists and the expensive microscopes conventionally used in diagnosis.
Our goal with ParasiteID is to enable more people suffering from parasitic infection in the developing world to be diagnosed and treated faster.
What inspired the project?
Kiersten: Almost 2 billion people — or 1 in 4 people worldwide — is infected with parasitic worms that cause neglected tropical diseases (NTDs). Infection with parasitic worms is among the most prevalent afflictions of people who live in areas of poverty in the developing world. These infections kill at least half a million people annually, but also severely decrease quality of life because they cause chronic sickness.
Through discussions with medical professionals, I was steered towards tackling a problem in the developing world where this kind of technology could have a bigger impact. Of course, the diagnosis and treatment of neglected tropical diseases is a problem of staggering proportions. We focused our Capstone on developing a diagnostic aid for parasitic worm infection because this problem was very amenable to computer vision solutions and a tremendous challenge in public health.
We were really inspired to help develop a faster way to diagnose and treat so many people and to help disrupt one of the underlying causes of the cycle of poverty in the developing world.
How did you work as a team?
Vicki: I was astounded by just how well our team worked together. We became a team because we were all excited and motivated by Kiersten’s vision for the project. Ultimately, we were united around a mission we all truly believed mattered — even beyond the Capstone project.
Fortunately, we all had interests and experience that complemented each other’s so in the end we were able to cover the main technical components needed in order to get ParasiteID up and running. We would never have had such a great outcome if we had each just worked individually on one component of the project.
Overall, consistent and positive communication and the hard work and ingenuity of all team members is what allowed us to produce the project we’re so proud of.
Overall, consistent and positive communication and the hard work and ingenuity of all team members is what allowed us to produce the project we’re so proud of. Beyond that, working in such a committed, positive, and supportive team made this Capstone project all the more meaningful and enjoyable. I learned so much from each of my teammates and look forward to continuing to work with them in the future!
Cameron: Because we had such high energy throughout the project, each of us took on less glamorous roles at some point. This was crucial to our success — we couldn’t all jump into the modeling! Also, because we discussed our progress frequently, nobody was stuck for too long without somebody else helping to get through roadblocks.
How did your I School curriculum help you prepare for this project?
Cameron: Our W210 (Capstone) instructors, Joyce Shen and Dr. David Steier, helped us shape a vision of our product and how to present it. Receiving continuous feedback from classmates and instructors helped us to stay on track and prioritize tasks well.
W266 (Natural Language Processing with Deep Learning) had a strong focus on research. So while we used the deep learning concepts from its curriculum, the real value was in the skills to research new technologies, read papers, and decide which new concepts would be best to implement — especially because none of us had ever done any computer vision before. That’s how we ended up with the object detection part (finding eggs among stool debris) — lots of research!
While we used the deep learning concepts from its curriculum, the real value was in the skills to research new technologies, read papers, and decide which new concepts would be best to implement…
W205 (Fundamentals of Data Engineering) taught us best practices in designing product architecture. Having a streamlined backend allowed us to make updates quickly and easily, and focus on the dataset and model.
W201 (Research Designs and Applications for Data and Analysis) helped us to present the idea in a way that communicated impact.
Vicki: 251 (Deep Learning in the Cloud and at the Edge) helped immensely — and not just because of the ongoing SoftLayer credits we put to good use for training and serving our model! The course was foundational for building intuition on how to design solid and secure system architectures in the cloud, which was imperative for our project. Also, Professor Brad DesAulniers gave us good advice and resources for how to go about adding bounding boxes to our images, which ended up being crucial for our model.
Do you have any future plans for this project?
Cameron: Yes! In some fashion, we would love to see the app used so that more people can be treated for these terrible, prevalent diseases.
The MIDS curriculum gave us the technical skills required to develop ParasiteID and familiarized us with agile product development and the importance of product-market fit.
Kiersten: We are in the very early stages of commercialization. The MIDS curriculum gave us the technical skills required to develop ParasiteID and familiarized us with agile product development and the importance of product-market fit. The next phase for getting ParasiteID into the hands of those who need it most will involve continued evaluation of product-market fit as well as tackling the logistics of product delivery. Fortunately, members of the MIDS alumni network have already reached out with support.
How do you hope this project will make an impact?
Vicki: If more people and organizations already working in the area of diagnosis and treatment of parasitic worm infection in the developing world were to use ParasiteID, we believe that it would allow more people to be diagnosed faster, and thus speed up their treatment and recovery processes. Furthermore, since infected individuals often re-contaminate soil and water supplies serving their communities, parasitic worm infections are public health problems, and ParasiteID’s role in speeding up diagnosis could, in effect, reduce this massive public health issue in the developing world.
Final thoughts?
Cameron: Please check out the project at https://parasite.id or parasiteid.com and try it out. If you have any ideas, connections, or resources we can use to move forward as an organization, please contact us via the contact page.
