Hospitals and clinics in South Sudan supported by Doctors Without Borders/Médecins Sans Frontières (MSF) are piloting the use of a snake identification database created with the help of artificial intelligence (AI) and machine learning. Dr. Gabriel Alcoba, MSF medical advisor for snakebites and neglected tropical diseases, shares his experience working with this technology and how it can benefit patient care.
I remember a time when we used photo albums to identify snakes in MSF hospitals. Medical staff would flip through pictures to figure out which snake had bitten a patient. I witnessed this firsthand in South Sudan and other countries while working on snakebite envenoming and other neglected tropical diseases.
Today, experts have created a snake identification database with 380,000 photos of snakes from different countries, using AI and machine learning. This innovative approach is now being piloted in two MSF hospitals in Twic and Abyei towns in South Sudan.
Together with the University of Geneva’s One Health Unit and support from MSF teams on the ground in South Sudan, as well as the medical and innovation departments at MSF in Switzerland, we are improving our database. We are developing software that uses AI to help identify snake species in the field, distinguishing between venomous and harmless snakes. This software can recommend the best action for a person bitten by a snake before they reach a hospital.
We are currently working on collecting quality photographs to feed into this software. South Sudan has one of the lowest numbers of snake ecological studies but also experiences high rates of snakebite admissions in MSF hospitals, especially from May to October.
What is snakebite?
Snakebite envenoming is a global health concern, associated with population displacement and climate shocks such as floods. Every year, more than 5 million people are bitten by snakes; of those, 2 million suffer severe effects, and between 81,000 and 138,000 die as a result of snakebite envenoming, according to World Health Organization (WHO) data.
Snake venom can cause three major syndromes: skin swelling, muscle destruction, and fatal hemorrhages or respiratory paralysis. Snakebite envenoming is classified as a neglected tropical disease, mainly affecting people in remote, rural, flooded, or conflict areas in low- and middle-income countries. Consequently, it is not a lucrative market for big pharmaceutical companies or policymakers, despite the millions affected, and research on effective, less complex, safer, and affordable antivenoms has not been prioritized.
Learn more about snakebite
Risky data collection process
For the AI to identify a particular snake, we need a photo of it, either while it is still alive or when it is dead. If someone has been bitten, the bite victim or someone nearby can try to take a photo of the snake after the bite occurs, but this must be done with extreme caution. If it was not possible to take a photo at the time of the bite, our staff can go back to the site of the bite and try to take a photo of the snake, again with great care ... When the snake is dead, [the photographer] should still take great care since it could still transmit venom.
Once we have a photo, it goes into the AI software, which compares it with thousands of images to identify the snake or add it as a new entry with GPS coordinates.
Early results are promising. The AI sometimes identifies snakes even better than experts! For instance, it can differentiate between venomous snakes like the Egyptian cobra or black mamba and harmless ones like the African house snake. With better-quality photos, funding, and more research, this snake AI app could provide real-time help for patients, from snake identification to choosing the right antivenom.
Snakebite: Seven bites of inequity
Snakebite is a neglected tropical disease that has received very little international attention. Here are seven key inequities at play.
More research is still needed. Often, patients receive the wrong treatment because the snake isn’t correctly identified, or valuable antivenom is wasted on bites from non-venomous snakes, which can also cause serious side effects. Antivenom is rare and extremely expensive, costing a patient anywhere from a month to a year’s salary.
The community-based Snakebite Awareness and AI Identification in Communities (SNAICS) project, along with the AI snake app, aim to improve our understanding of different snake species and optimize our antivenom supply, ensuring effective clinical responses.
