"Taking AI to the next step": Q&A with Arsen Muhumuza
Arsen Muhumuza, a student in his final year of medical school at the University of Rwanda, became involved in AI and health through his participation in a 2017 hackathon that helped him generate ideas for how to leverage technology to address health challenges. Since then, Arsen has helped organize multiple events with the goal of convening a range of stakeholders and innovators to develop surgical solutions using technology. At his university, he has founded the Rwanda Digital Initiative and hosted a boot camp for people to focus on different health challenges and AI-supported solutions. We conducted an interview to hear more about how he got here and what matters most in AI and health. These responses have been edited for length and clarity.
Kate Raphael: What are the most pressing health issues that can be potentially addressed by AI and data science in Africa, and specifically in Rwanda where you work?
Arsen Muhumuza: We have very serious problems with the number of doctors and the number of patients. The balance is very disturbing. One doctor sees a large number of patients and that’s a huge gap to cover without AI systems. There’s an additional problem when it comes to the software used to keep patient data. Not only are doctors seeing that many patients, but they have to document all of that information in the system; we need something that solves that problem.
Each day, I receive 20 patients. On some days, it’s more than 50. There are so many people traveling long distances to find a doctor. And given the number of surgeons, and the number of patients who need surgery, we need to think about what AI could do to solve that problem. Like so many things, when I look at my country, we need something to boost the kind of services we provide to people. I think we are a little bit behind.
What challenges do you anticipate in trying to address these specific health problems with AI?
One of the challenges is that most of the solutions available are not homegrown. Developments from different countries aren’t necessarily well-suited to our needs and systems. If something has been developed in our region, there’s a greater possibility it will be very helpful. The people who are supposed to be using new technologies are somewhat resistant to changing, and some don’t understand how to use it. Most of the solutions we try to implement are very important, but the government is very hesitant, and AI is not one of their core priorities.
What are the “right approaches” for integrating AI into clinical care? What could we be doing, or what aren’t we doing that we should be doing?
One of the most effective approaches is leveraging individual institutions, like universities and hospitals, to advocate for the advancement of a potential solution. One of the reasons for this is because when you look at our system, potential solutions have to be tested, and if they go to the government first, the process is really complicated, and most of the government officials have no idea what these proposed tech solutions are or why they’re important.
When solutions come from individual institutions, they have more a voice. So, for example, if it’s an AI solution that has been developed at Harvard and there’s a project to deploy it in East Africa, one of the best approaches would be to work with universities and health care institutions to get these solutions into the teaching hospitals. Then they’re more likely to advocate for these solutions for the rest of the country. However, if you bring the solution over via the government, then they are much less likely to implement it.
How should we develop standards for how AI should be used effectively and ethically in medicine?
Usually when we try to develop something that’s going to be used by the population, it’s very important that there’s an ethical way of doing these things. If it’s going to affect someone’s life, it has to be developed in a sensible way. With respect to AI solutions, someone might develop a solution that hinges on collecting patient data, and it’s meant to solve patient problems, but we have to be careful that it’s never misused. Any solution should be checked and reviewed before being implemented everywhere. Everything that has to do with patients and communities necessitates ethical approval.
Who do we need to partner and collaborate with to better address health challenges using AI? Who is often at the table, and who do we need to engage who isn’t often in the conversation?
We should be talking with people who can help develop solutions, but also the people who have a mindset of taking AI to the next step: the generation that will roll with these ideas and keep pushing AI solutions. In our settings in Rwanda, where we’re still somewhat behind in the field of AI, we need to get the universities on board. They are specifically doing research and training the next generation of health leaders, so involving the universities is the first step. The next step is involving NGOs who are trying to develop the health of communities. Then, doctors and hospitals have to be involved. If you develop an AI solution without doctors’ help, you have to change their perspective about how to use it; you’ll bring the AI solution in, but no one will use it. So, doctors are crucial for developing the AI solution itself.
There are other people that should be involved and are not already involved, like businesspeople. There’s a difference between an AI entrepreneur who already has an AI solution and a businessperson who wants to invest in something. We should find people to invest in AI who can really buy into the ideas.
We also cannot forget the government. Some countries’ governments are able to bring in something new that really works. They should be on board when you have different solutions.
What do you wish we were focusing on more in the realm of AI and health?
Data collection and analysis should be the primary focus. To have a good AI solution, you have to have the data.
Advancing research for AI should be secondary. Then we have to develop an AI solution and we have to use research solutions to do that. I think AI is very important in diagnostics in reducing the number of errors by health care professionals. That addresses a huge problem.
My country would also say we need AI to reduce the patient-doctor gap. We don’t have enough radiologists to do diagnostics, and the ones we do have, they have too many patients. We don’t have enough people to do laboratories or run exams. Sometimes it takes days or weeks to get a CT scan. It may not cover the doctor-patient gap, but we can alleviate it a little bit. Some things can’t be done with just doctors.
Finally, we need to invest in ideas that work in different countries. For example, in organizing this event, most of the people invited are already developing AI solutions, or already have a background in tech and health care. We have to invest in some ideas to really work on problems that desperately need to be addressed. For most of the AI solutions addressing problems in other countries, if that technology is brought to our country, it won’t have the same effect. We have good ideas here, but we don’t often have enough funding to produce something from those ideas. But, when you have good ideas, and good support, then you can develop something great.