Artificial Intelligence (AI) is already changing healthcare as we know it. When it comes to identifying numerous life-changing diseases, studies indicate that deep learning or machine-based algorithms beat medical professionals. For example, one international dermatology study, spanning Germany, France and the US, used a deep learning neural network to identify skin cancer. It fed the network more than 100,000 images of malignant melanomas (cancerous skin patches) and benign moles. After the training, the algorithm accurately detected 95% of the melanomas compared to 58 international dermatologists with an 86% detection rate.
Research studies like this showcase the potential for AI to elevate the work of medical professionals. AI algorithms run on large amounts of medical data and use machine learning to identify patterns and make predictions. Instant access to vast amounts of medical information allows AI to increase the accuracy and speed of diagnosis and assist medical professionals in prescribing better treatment.
According to experts, the speed and precision of scanning algorithms work best in conjunction with human experience. It is crucial to be aware of the shortcomings of AI, especially when considering highly technical analysis. One study detecting breast cancer in patients found that 36 AI results were 94% less accurate than one radiologist and 100% less accurate than two radiologists. It is an indication that currently, AI is more suited to perform less specialised tasks such as recording medical paperwork and reviewing less technical scans. Studies in the West show that doctors spend up to twice as many hours on these tasks compared to face time with their patients. Automating such mundane tasks would free up doctors’ time to focus on the more specific and critical diagnoses. In the near future, it seems AI will not replace humans yet will speed up diagnosis allowing more efficient use of the time of medical professionals. Further down the line, as AI developments advance, diagnoses that previously required humans could be completed by an algorithm. Such advancements will boost doctors’ productivity and effectiveness.
With early detection algorithms becoming more reliable and widely available, this also opens huge potential within developing nations for AI to increase the distribution of healthcare. The lack of supply for adequate healthcare in developing countries is illustrated best by the number of doctors per capita being significantly lower than in the West. South-eastern Asia and Sub-Saharan Africa have only 6.69 and 2.1 doctors per 10,000 people, respectively, whilst Europe has 47.24. Health groups such as Babylon have picked up on this problem and the opportunity for AI to provide a solution and began pilot operations in Rwanda in 2016. Since then, its telemedicine service has had 2 million users register across numerous African nations and handles 3,500 daily consultations. Its app-based service provides a simple to use AI-driven symptom checker, active monitoring of health metrics and the ability to talk to a clinician virtually. Similar projects such as Ada Health’s chatbot symptom checker app with 9 million users also utilise AI-based questioning in lower-income countries to pinpoint possible diagnoses and direct users to available treatment. Solutions like this, whilst not perfect, prove that simple AI can relieve pressure when it comes to diagnosis in developing nations healthcare systems. This technology, in turn, allows channelling existing healthcare resources towards infrastructure and treatment. It will enable medical professionals to focus on treating more patients as AI picks up the slack of diagnosis.
AI projects like this in developing nations are broadening the reaches of healthcare to every corner of the world. However, there are still significant constraints and problems that need to be solved moving forward. AI relies on technology that is not nearly as advanced and accessible in developing nations compared to developed countries. Projects such as Ada Health’s digital diagnosis relies on patients having access to smartphones and an internet connection, which is not a given in many developing countries. Investment in basic infrastructure across all nations needs to occur before the rewards of AI can be fully reaped internationally.
There certainly is scope for AI to broaden global healthcare distribution and the examples outlined are already demonstrating the potential of AI in this field. However, it is necessary to appreciate the constraints of AI at this point. We must understand the detection accuracy limitations of AI and acknowledge the substantial technological infrastructure required for AI diagnosis to revolutionise healthcare distribution across the world.
By Alex Young
Sector Head: Anuar Gaisin