As it looks to curb the spread of infectious diseases, the Gates Foundation is lending its support to an effort to use artificial intelligence to make it easier for health providers in India and Nigeria to quickly diagnose certain conditions.
The funding will support New York-based clinical decision support company VisualDx, which began as a collection of images to allow non-dermatologists to better identify skin conditions. The company now offers machine learning-powered apps and software platforms to help health care workers quickly and accurately diagnose patients.
“The public health needs of many rural and underserved areas often go unmet from provider and doctor shortages, and limited access to diagnostics, assistive technology, education, and training,” said Wendemagegn Enbiale, global health information officer of VisualDx, in a press release. The company’s hope is that their tool can help health care workers in places where there aren’t enough specialists.
The foundation previously provided financial support for VisualDx to develop an offline version of its tool in Botswana. In the new program, the company plans to work with partners in Nigeria and India to both gather new images of neglected tropical diseases such as trachoma, hookworm, and echinocccosis, among other diseases.
“It’s not good enough just to say to a health care worker, ‘Well, here’s an app that shows you the rare infectious disease,’” said Art Papier, Visual Dx CEO and a dermatology professor at the University of Rochester. Providers will need to be able to quickly parse whether the lesion is due to a common skin condition that won’t cause further harm, or whether it’s a sign of a neglected tropical disease that warrants a specific course of action.
Skin disorders can also present much more subtly on darker-skinned people, who have historically been left out of medical literature. Whereas the inflammation from a rash is easy to visualize as red against white skin, the increased red blood flow on brown skin doesn’t look red, said Papier, but instead looks like a deep brown. “It’s very, very important that you show how disease looks and people of all skin types because it looks very, very different across skin tones,” he said.
The software compiles images of the patient’s skin with other factors, such as symptoms and if the patient has been traveling, to help suggest diagnoses. Importantly, Papier said that the AI does not make a decision on the diagnosis, which is medically difficult to begin with.
“We’re talking about people’s lives. I mean, you can’t just wing it,” he said. “You’re saying, ‘this is how the clinical decision support engine has interpreted all the clues and these are the things that you should be thinking about,’ but the human makes the final decision.”
Communicating the limitations of the tool to the health care providers using it will be crucial if VisualDx wants the system to prove useful in the field. So, too, will be building trust with the providers in the program.
The company will also need to ensure its system translates to a variety of health care settings in different countries and works well with a range of patient populations, an issue that has long impacted the performance of machine learning models in health care. Papier said the funding will also help VisualDx develop “country-specific logic” for the tool based on the prevalence of the disease in that part of the world.
Diagnosing these diseases won’t address more systemic issues that drive their spread, such as the prevalence of mosquitoes and black flies, and the lack of access to clean water and sanitation. Still, the ultimate hope is that better diagnosis can lead to better treatment rates for neglected tropical diseases — particularly those that can be controlled with medicines or other interventions.
As climate change increases, so also has the spread of infectious diseases, including zoonotic and tropical diseases. “These diseases are on the move,” Papier said. “And whether Zika virus or other viruses we have, the only sensible thing is to prepare for everything.”