Join the online conversation on deep learning for healthcare at the BoP here!
Luke, who lives in London, will celebrate his fifth birthday tomorrow; a celebration, but not his most life changing occasion. For the parents of Akachi from Mafubira (Uganda), however, his fifth birthday is momentous: Akachi had more than thirteen times higher chances of being amongst the under five year olds that die prematurely from preventable diseases. These numbers seem unimaginable considering today’s medical advances; but millions of people continue to lack access to medical facilities, physicians and medicine to the detriment of their health.
Imagine if these shortages in doctors could be alleviated by robots?
Artificial intelligence (AI) could make this a reality. During the industrial revolution, machines made human muscles a thousand times stronger. The big data revolution will make the human brain a thousand times more powerful. In healthcare, AI has the potential to augment the expertise of trained healthcare personnel and transform the way we think about diagnosis and treatment of diseases. Already now, algorithms are improving early detection of diseases, the development of treatment protocols and patient monitoring and care. Studies point to its potential to improve health outcomes by 30-40%, and to cut costs of treatment by up to 50%.
So what exactly can Artificial Intelligence do for patients?
AI refers to algorithms and software used to analyse complex medical data. For example, Philips has developed a solution for automatic detection of tuberculosis (TB) from chest X-rays in combination with various other data points. TB is still a leading cause of death South Africa, often due to delayed diagnosis. The solution is particularly helpful in contexts where radiologists are scarce, also bringing down costs.
Google is currently working on an AI solution that can identify diabetic retinopathy in retinal photos using deep learning. The disease is a leading cause of blindness amongst diabetic patients. The project at Google was initiated by a researcher who realized that in his native India, most people are not screened for the disease due to a lack of capacity.
Sounds interesting but…how can this help patients in underserved regions?
Whilst this technology is currently being developed within the global North predominantly, the potential for impact is significantly stronger for remote areas with no access to physicians. Take the example of AI chatbots such as the one from Ada. Such chatbots allow individuals to take control of their personal health. They significantly reduce the burden on medical professionals regarding easily diagnosable health concerns, and can also reduce medical error. In Germany or the UK, this can help patients save the time and hassle to visit a doctor for minor issues – and empower them to manage their own health. In rural Kenya or Tanzania, such solutions could augment the expertise of healthcare workers where there is no doctor or specialist available.
Babylon Health is doing exactly that: The company’s apps offers video consultations with real life GPs and specialists. In Rwanda, 100,000 citizens have already signed up to Babylon as part of a subsidized government scheme to increase the reach of the country’s health services. The company is also starting to offer artificially intelligent services for triaging patients.
However, applying AI technology to achieve better health outcomes in underserved regions will not come without challenges, including…
• Acceptance – AI cannot replace compassion; there is no algorithm for empathy. Particularly in countries where strong relationships with traditional healers persist, it may be a challenge for patients to trust in the technology. Similarly, doctor receptivity may be a challenge. Whilst AI will not replace the need for doctors, it may change their way of working and replace some tasks that they frequently engage in, like screening for disease. Most certainly however, it will help physicians to become ever more efficient and accurate in their diagnosis and treatment.
- Availability – While AI may be able to alleviate the shortage of doctors, the many barriers to proper diagnosis and treatment may still persist. In fact, applying AI is useless if there is no follow up treatment options, or if medicine is unaffordable. Also, improving internet connectivity is an essential prerequisite for making AI services available. In 2017 only 42.9% of households in the developing world have internet access.
- Affordability: While the technology is promising, it does not come without costs. At the bare minimum, patients who want to use AI chatbots need a smartphone – excluding many from the service who don’t. In addition, the availability of diagnostic devices to inform AI solutions with data may not be available or examples may simply be unaffordable.
- Navigating regulation/ethical boundaries: Dehumanizing healthcare raises ethical issues regarding information, data management and security, legal liabilities and the question of who ultimately controls this technology. In many countries, however, regulation does not yet exist, leaving patients without protection. Regulators will need to strike a balance between incentivizing innovation and protecting the vulnerable.
…however, many actors are already working on overcoming these barriers:
In 2015, the social enterprise Living Goods launched a trial: They offered community health promoters a smartphone with an app that would help them to better register, diagnose and treat children and pregnant mothers. Along with higher rates of self-confidence and technological knowledge, today these apps are central to ensuring community health in rural Uganda.
Now imagine the impact such app-based services could have on the health of underserved patients if they were enhanced with artificial intelligence? Additionally, thinking one step further, how would community health look like if AI apps were paired with portable and affordable diagnostic devices such as those from I-nside? The company offers a portable endoscope, which can easily be attached to a smartphone camera, to diagnose ear related illnesses via photo analysis.
What if such health data, generated by community health workers, could be stored in electronic patient records that are owned by the patient, and only accessible through an iris scan? Well, UNUMED has just piloted this solution in refugee camps in Kenya, allowing them to take health data with them wherever they go.
Looking ahead: the future is closer than you think!
In contrast to what popular science fiction portrays; we are not five years away from a futuristic, machine driven world. AI will never replace the importance of human interaction. It will, however, work alongside traditional healthcare structures in order to ensure that children like Akachi can celebrate their birthdays free from the fear that they will become just another statistic.
Want to be part of the solutions? Get engaged!
Are you working on artificially intelligent solutions in healthcare? Have you invented affordable, mobile diagnostic devices? Are you an expert on data management, blockchain technologies and storage? Or are you a social entrepreneur or NGO improving community health in the remote areas of this world? If yes, our upcoming event ii2030 is for you! ii2030 we aims to bring the right experts together to develop the inclusive innovations that will revolutionize healthcare in the underserved regions of this world. Join us on October 18 and 19 in Berlin – or join the online conversation atiba.ventures!
About the author: Dr. Aline Menden is a Founder and Managing Director of Endeva, and hosts the track “Intelligent Diagnosis and Treatment” at ii2030. The track is sponsored by Bayer. Aline would like to thank Stephanie Rogers and David Jarry for their contributions to this blogpost!