BBC News Arabic on how How AI is Revolutionising Healthcare

BBC News Arabic service interviewed INSIGHT director Pearse Keane for a recent feature on AI in healthcare. With kind assistance from BBC reporter Somaya Nasr, we are able to reproduce below an English translation of the original Arabic article, which can be accessed on the BBC website here.

What was once considered science fiction is now becoming a growing part of modern medicine: artificial intelligence has already begun transforming healthcare across the world, with applications ranging from disease detection and drug development to hospital management and remote patient monitoring.

Healthcare systems, universities, and research bodies in the United States and Europe, as well as several initiatives in Arab countries, are increasingly using these tools to help doctors diagnose diseases more quickly and improve patient care.

Many experts believe the technology could represent a major breakthrough in medicine and help ease pressure on overstretched healthcare systems. However, concerns and questions remain regarding safety, regulation, and bias.

Excerpt from the BBC News Arabic article

Key achievements so far

“There are two main areas where AI is starting to make a difference in healthcare. One is large-language models such as scribes, which help doctors with note-taking and translation, making it easier to communicate with patients”, according to Professor Pearse Keane, consultant ophthalmologist at Moorfields Eye Hospital in London and Professor of Artificial Medical Intelligence at the UCL Institute of Ophthalmology.

The second area is medical imaging analysis. Professor Keane told BBC Arabic: “In any specialty where you have large amounts of imaging data such as ophthalmology, radiology and neurology, there are opportunities to develop AI models trained on more images than a human expert may ever see in their lifetime. These models perform as decision support tools, capable of analysing and interpreting a patient’s scan to help the clinician reach an informed decision about treatment.”

He added that “medical imaging AI is being used in stroke units throughout the NHS, and in many cancer units. The evidence so far is that diagnostic errors have been significantly reduced. In ophthalmology there are around a dozen AI models approved for clinical use, a few of which are currently being piloted in the NHS.”

Moorfields Eye Hospital, in collaboration with Google-owned AI company DeepMind, has also begun deploying an AI system capable of analysing retinal scans and identifying diseases requiring urgent treatment.

In the UK as well, a recent study conducted by the NHS in collaboration with Imperial College London and Google found that, compared with human assessment, the use of AI led to the detection of more advanced and overall cancer cases, fewer false positives, and fewer women being recalled after initial screening, while also reducing the time required to analyse scans by about one third.

In the United States, researchers at the Mayo Clinic — a leading medical research institution and hospital nationally and globally in healthcare — have developed AI systems capable of identifying signs of heart disease from routine electrocardiograms (ECGs), sometimes even before patients show clear symptoms. Meanwhile, China has established the world’s first AI-powered hospital and has deployed medical large language models across hundreds of hospitals to assist in diagnosing rare diseases and identifying critical cases.

Arab efforts to deploy AI in healthcare

While efforts to deploy AI in healthcare are largely concentrated in Europe, the United States, and China, many Arab countries — particularly in the Gulf — have begun investing in integrating this technology into their health systems.

For example, Saudi Arabia has launched the “Seha” hospital, which uses AI tools to connect dozens of hospitals across the country and provides telemedicine services to support the healthcare system, improve access to urgent care, and help doctors prioritise critical cases and suggest treatments.

The UAE has recently ranked 15th globally in the Global AI Competitiveness Index published by the Deep Knowledge Group for its applications of artificial intelligence in Healthcare, Biotech, and Longevity. Abu Dhabi’s Department of Health also launched the “Population Health Intelligence” platform last October, which uses AI to manage population health. The system is based on a shift towards a proactive healthcare model that includes disease prediction, with an initial focus on tackling obesity and enabling early cancer detection.

Farah Shamout, Assistant Professor at NYU Abu Dhabi and Lead of Clinical AI Lab, told BBC News Arabic that the Gulf region is well positioned to become a leader in adopting this technology in healthcare.

She added that “Many hospitals in the Gulf already operate state-of-the-art healthcare systems that generate high-quality clinical data, which is essential for the development, evaluation, and safe deployment of AI models. At the same time, governments and healthcare institutions are investing in understanding the regulatory frameworks and risks associated with AI adoption. In my view, this is the most critical step in such a high-stakes environment, where patient safety and quality of care must remain the top priority.”

Other emerging initiatives exist across the wider Arab world outside the Gulf. In Morocco, for instance, AI is increasingly being integrated into healthcare, with hospitals using the technology for medical image analysis, disease diagnosis, and expanding telemedicine services.

In Jordan, the “Virtual Hospital” was launched last year, connecting several remote hospitals to a central command centre. It uses AI technologies to monitor patients remotely, analyse radiology images and transmit them to specialists, triage cases based on severity, and enable continuous monitoring without requiring patient transfers to the capital.

Constraints and challenges

On the biggest barriers preventing wider adoption of AI tools across advanced healthcare systems such as the NHS, Professor Keane says they are complex in nature.

He explains that these factors range from “fragmented data infrastructure across different parts of the NHS to lack of confidence on the part of clinicians,” as well as the need to develop new approval frameworks that must be continuously reassessed as the technology evolves.

Professor Keane also notes that “if that means AI adoption takes a little longer it is not necessarily a bad thing.”

Several Arab countries outside the Gulf face a different set of constraints, largely due to limited healthcare resources and digital infrastructure.

Dr Omayma Omari Harakat, a France-based researcher in management sciences, specializing in public management and healthcare systems who authored a paper titled Artificial Intelligence in Morocco’s Healthcare Sector: Ethical Challenges and a Framework of Responsibility, says these obstacles are multiple and interconnected. They include:

• Infrastructure: Many health facilities still require stronger internet connectivity, interoperable health information systems, electronic medical records, secure data storage, and reliable medical equipment.

• Human capacity: Doctors, nurses, and hospital managers need training not only in using AI tools, but also in evaluating them — including questioning their training data and limitations.

• Regulation and trust: Patients and professionals need clarity on how health data is protected, who is responsible in case of errors, and whether private companies can access or transfer medical data.

• Financing: AI adoption is not limited to purchasing software; it also requires maintenance, data governance, cybersecurity, training, evaluation, and integration into clinical workflows. Without sustainable funding, pilot projects risk remaining isolated experiments.

Bias risks

Experts emphasize the importance of developing region- or country-specific datasets rather than relying exclusively on Western medical data.

Shamout explains: “This is what we often refer to as “data distribution shift” - when an AI model is trained on one population but deployed in another with different demographic, genetic, cultural, or healthcare characteristics. Such differences can significantly affect model performance and reliability.”

She stresses the importance of “building region-specific datasets and strengthening local research capacity” to ensure that AI systems are clinically accurate, equitable, and clinically relevant for Arab populations.

For example, in dermatology, AI systems trained primarily on lighter skin tones may be less accurate in diagnosing conditions on darker skin, according to Dr Omayma, who notes that “this is not a minor technical issue; it can produce delayed diagnoses and reinforce existing health inequalities.”

Professor Keane also highlights the importance of training medical AI models on diverse datasets to ensure “work equally well on a wide range of people.”

He adds that this is a priority for his research team at Moorfields Eye Hospital and the UCL Institute of Ophthalmology: “we are taking this a step further and developing the world’s first globally representative medical AI model, trained on more than 100 million eye images gathered from people in more than 70 countries across every continent except Antarctica. We will make this model freely available for non-commercial research to advance the development of more equitable AI in healthcare”

Regulation, trust and public perception

As AI tools continue to evolve rapidly and expand across sectors, experts say regulatory frameworks are struggling to keep pace.

Healthcare in particular requires specific rules governing medical data use, algorithm approval, clinical validation, cybersecurity, patient consent, and legal responsibility, according to Dr Omayma, who believes that “the region is moving, but healthcare AI governance must become more specific, more operational and more anticipatory.”

For Dr Omayma, the challenge facing Morocco and Arab countries more broadly is avoiding two opposing extremes: “blocking innovation through excessive bureaucracy, or allowing unregulated deployment that could expose patients to risks”

Another important element is the trust of both clinicians and patients in AI models, as well as how society perceives them.

Farah Shamout says that “public trust and cultural relevance will determine whether both patients and clinicians feel comfortable using AI systems in practice. For example, in our own work, we have examined the performance of AI models on Arabic medical tasks and found significant disparities compared with English-language performance. This has already affected some products in the market”.

She adds that “solutions that are not adequately adapted to local language and cultural contexts are unlikely to gain widespread acceptance or deliver reliable outcomes.”

One key question remains, and it is likely to come to the minds of many readers: Will artificial intelligence remain a tool that supports doctors, reduces pressure on overstretched healthcare systems, and improves patient care and access — or will it fundamentally reshape the practice of medicine to the point of reducing the role of the human clinician?

Most experts agree that the human element will remain essential, particularly in complex diagnoses, ethical decision-making, and patient communication.

“It is unlikely that medical AI will ever replace clinicians,” Professor Keane said. “Studies show that the optimum experience and outcome for patients is having a human as part of the care process, not leaving it to AI alone.”

Many believe that AI could bring a major transformation to healthcare. However, its success will depend not only on technological advancement, but also on clear and continuously updated regulatory frameworks, alongside public trust and human oversight that ensure it is used safely and fairly in a way that benefits patients and relieves pressure on healthcare workers.

---