Predictive healthcare: how AI can save lives by identifying diseases faster
 |
As we approach COP28, which will focus on healthcare as part of its sustainability agenda, it is critical to assess the challenges that the globe faces in providing healthcare services to a growing global population. According to the World Health Organization (WHO), up to 3.5 billion people—over half of the world's population—lack access to essential healthcare services.
The causes of this disparity in healthcare service distribution are complicated and multidimensional, and correcting it will require herculean efforts by governments, research organizations, and the public and private sectors. Regardless of how difficult this mission appears to be, I am quite hopeful about the role of AI, which I believe will be critical to improving access to healthcare for traditionally underprivileged populations.
I previously worked with a team of researchers as a machine learning specialist focused on bringing AI to difficulties in medical image processing to develop a solution to help healthcare practitioners measure fetal growth at a pace and precision that would not otherwise be feasible.
At MBZUAI, we are in the early phases of investigating how machine learning might be used to detect cardiovascular abnormalities in CT scans and possibly forecast future consequences. To that end, we are conducting a massive study in partnership with the University of Oxford to identify biomarkers that could anticipate potential heart problems before they appear as more visible symptoms. The Oxford team has already acquired 150,000 CT scans and long-term prospective results from patients all across the world. We will then train the AI system to distinguish between healthy cardiovascular systems and those that have—or are likely to develop—problems.
This large Oxford CT scan study is exciting, but I also want to go further and make sure that comparable transformative technology reaches people in remote and developing areas, where challenges like cardiovascular disease and fetal anomalies frequently go undetected with devastating results. However, broadening the solution in this way would pose certain special difficulties, such as how to conduct scans on individuals without access to a cutting-edge clinic or hospital. This is where the development of new technologies, such as portable ultrasound devices and the ability to push artificial intelligence (AI) to the edge of the network, holds the promise of saving many lives.
We are attempting to employ ultrasound scans for predictive diagnosis with my MBZUAI research team. Because it might allow medical personnel to scan patients' hearts and monitor the AI system's interpretation of these scans, the potential advantages of merging portable ultrasound scanners with AI in developing nations are increased. This recently discovered information might draw attention to important issues that require more study.
It is important to note that the AI component of this medical procedure aids the human physician rather than replacing them by performing the labor-intensive task of evaluating images and highlighting those that require attention. This program could provide millions of individuals with the first access to dependable cardiovascular exams in many parts of the world.
Analyzing foetal growth and spotting anomalies follow the same logic. We intend to carry out research focused on employing portable ultrasound scanners to carry out the kind of 'anomaly scan' that is common in wealthy nations but carried out on a much lesser scale in less developed regions and can be regarded as uncommon in communities in most of Africa and Southeast Asia. We will need to train the AI system to recognize the fetuses that require further attention and learn how to measure their growth, just like with cardiovascular imaging. The beauty of such a system is that nurses or midwives might administer it to smaller clinics.
These scans are crucial because early detection allows for the treatment of anomalies such as congenital heart problems or spina bifida during pregnancy or shortly after birth. Additionally, the time it takes to evaluate each scan might be cut in half by using AI to find anomalies, from a mean average of between 40 minutes and one hour to just 10 to 20 minutes.
All of this is still in the very early stages, but I am confident that, together with my talented team at MBZUAI, we can do our part to improve healthcare for the millions of people who are now underserved.
The future is already here; it's just not divided equally yet, as American author William Gibson once stated. At MBZUAI, we are committed to extending the benefits of science, and in particular AI, to everyone on the planet.