Professor Ibrahim Adeola Katibi, a renowned cardiologist, teaches at the University of Ilorin in Kwara State. He is a former Dean of the Medical School and the current director of the Central Research Laboratory at the institution. In this interview with ABDULLAHI OLESIN, Katibi sheds more light on why he built artificial intelligence (AI) to read ECGs for African patients.
What first sparked the idea that Africans needed their own ECG-interpretation system. Was there a particular patient case that made the gap impossible to ignore?
There have been many instances of ECG machine misdiagnosis in hospitals, which are promptly corrected by the cardiologist over-reading the ECG. But of course, there are only about 600 cardiologists to 250 million Nigerians (a ratio of 1 cardiologist to about half a million people, as against the USA, where it is 6 per 100,000, or 1 to 10,000, as recommended by the WHO). Therefore, cardiologists would not always be available to every patient with a heart condition, hence the need for some degree of automation of diagnosis via artificial intelligence (AI).
To answer your question more tangentially, our team, over the last decade, has demonstrated incontrovertibly the sharp racial variations in the ECGs of Nigerian, British, Chinese, and Indian populations, and these are available in public domains for those who are interested in reading further. The situation is even better exemplified now that we have several Nigerians, and indeed Africans, working and living abroad these days as a result of the massive JAPA syndrome.
When you first shared this concept, did colleagues locally or abroad doubt the need for Africa-specific ECG data, if yes, how did you respond to that scepticism?
No, the need had never been in doubt, but who will bell the cat is the challenge because of the rigour and span of the work. The peculiarities of the African ECG have been well known over the years. The need to localise the solution had been established a long time ago. But unfortunately, even though computer application to the interpretation of ECG began in 1960, the pioneer work in Nigeria on computer analysis of the ECG was done by our team in the year 2010. By this, I mean the real application of computer programming to extract hundreds of raw measurements from an indigenously Nigerian ECG and interpret them through automation, as against importing Caucasian-calibrated ECG machines and using them on Nigerians, irrespective of errors of interpretation. Our colleagues, therefore, saw the need but were literally helpless due to a lack of equipment and/or technical know-how.
Our colleagues outside Nigeria were very eager to support our dream in order to compare it with what they already knew. The mistake we did not make was handing over clinical data to them and turning our backs without spearheading the evaluation and technological development. This endeavour has therefore evolved over the last two decades or thereabout, which again is the power of consistency and focus.
What are the most significant physiological differences in ECG patterns between Africans and Caucasians that the public should know about?
Very many of them. Perhaps the only similarity is in the PR and QT intervals, which I don’t know how the layman could understand. By and large, Black Africans tend to generate higher voltages than their White counterparts matched for age and sex. What this translates into is that when ECG machines calibrated for Caucasians are used in Africans, there will be false reports of chamber enlargement. Africans also tend to have ST-segment elevation as a normal variant compared to British and Chinese populations, which, when not interpreted in the proper context, would give an erroneous diagnosis of heart attack. Nigerians also tend to have a faster heart rate than the Chinese or British population, while the QRS duration is longer in the Chinese population. Hence, the possibility of using the ECG to identify the biological race of an individual at borders or in forensic medicine is also possible.
How do those differences affect diagnosis and treatment outcomes?
To start with, you need an accurate diagnosis of the problem before treatment can be effective. So, where ECG machines that were not calibrated with Nigerian or African data give false diagnoses, the treatment is bound to be defective, if not outrightly harmful or even fatal. In essence, the message is that there has to be local content in technology deployment or acquisition in order to avert inadvertent errors, which in health care could lead to catastrophic consequences.
How exactly does your AI learn from African ECG patterns? What data or methods make it uniquely accurate?
AI is essentially pattern recognition and drawing familiarity inference after getting used to the situation or object. This could be likened to introducing an elderly family friend to a small child over a period of time. The first two visits, the child would not come near the visitor, let alone allowing him or her to touch him. But he is studying him and noting the patterns of his behaviour. If the visitor is smart and comes along with sweet and toys, that facilitates the studying/acquittance and when the child has identified adequate patterns around the visitor, the child begins to see him or her as a friend and he begins to move towards the visitor and even allow the visitor to carry him and embrace him. Don’t be surprised, the child would soon begin to request; can we go to uncle lagbaja’s house please. The computer, through “Artificial Intelligence” can also be trained to recognise the pattern that it has seen before in a similar fashion and this is what AI is all about.
Our AI invention leverages real-time ECG data captured from machines using XML format. This data is automatically collected, parsed, and transformed into a structured format that can be analyzed by advanced machine learning algorithms, using supervised machine learning technique. The system uses robust, state-of-the-art techniques to predict potential heart abnormalities, empowering both patients and clinicians to take timely corrective actions. The software is tailored specifically for the African healthcare context to ensure cultural relevance and practical usability.
The result is a fast, intelligent heart condition prediction tool capable of delivering
accurate diagnostic insights in under three minutes. By integrating AI and cloud computing technologies, this system provides a scalable solution for early diagnosis and preventive healthcare. It was designed with the end user in mind, the software aims to reduce the incidence of undetected heart conditions by up to 85%, ultimately contributing to better health outcomes across underserved populations in Africa.
The strength of our research is that we have painstakingly collected a database of normal ECGs in Africans, particularly Nigerians over the years, such that we have the largest database of normal electrocardiogram (ECG) in any indigenous African population worldwide. Intentionally, we have also collected sizeable number of ECG datasets for heart failure and Angolan population.
What were the hardest technical challenges during training?
Ans: As a practising Cardiologist, Professor and researcher, it was not difficult for me to identify the gaps in knowledge and which of those gaps do I want to fill. In addition, I attend international conferences regularly as a Fellow of the Nigerian Cardiac Society, Fellow of the American College of Cardiology, Fellow of the European Society of Cardiology and Fellow of the Royal College of Physicians of UK.
The most important deterring factor would understandably be acquiring the ECG database to mine but again, for me, this was not a great deal because I have always had my eyes on the gold mine. The real technical challenge is assembling young men and women to be part of the team to pursue the idea to a logical end. The data scientists and machine learners are brilliant young men and women who are highly sought after and can work anywhere in the world remotely and still earn good, if not better income than we can offer them. One of the biggest challenges was leading such a team and keeping them together despite our meagre resources. In the first multidisciplinary team I assembled, we lost sixty percent of the researchers or team members to greener pasture outside the country within the first year. So keeping and sustaining such a highly sought-after team of multi-talented and highly skilful individuals was indeed a very big challenge.
Another big challenge is affording the cost of cloud storage and some of the Apps necessary for the analysis. These were paid for mostly in dollars but we earn income in Naira. But for the assistance of great philanthropists like E-Tech services under the Chairmanship of Mallam AU Mustapha, SAN, who gave not only of his resources to sustain the cost but also of his managerial experience and long reach.
Can you recall a real-life scenario where an imported machine misread an African patient’s ECG and your system corrected it? How common are these misinterpretations in your experience?
We see misinterpretation on a day to day basis, the gravity or consequence depending on the experience and versatility of the managing doctor. Smart non-Cardiologists have learnt to seek for the over-reading of machine-interpreted ECG before jumping into final action.
Personally, I have cautioned or stopped taking many patients to the cardiac labs or theatre just on the basis of false alarm by those imported machines for Nigerians living abroad when I re-assured that the finding under consideration is a mere normal African variant, thereby saving the risk and cost, not just to the individual patient but also to the system.
Your tool can work on a mobile phone. How does that change access to cardiac care in rural or underserved communities?
Access to cardiac care is likely going to be revolutionised through the deployment of this innovation. At the moment, there is one Cardiologist to about half a million Nigerian population. The figure is definitely worse in the rural areas since Cardiologists are extremely unlikely to reside in the rural area to practice. What we have done is to design a mobile App that would allow any individual, be it health care worker or layman to record and have their ECGs interpreted instantaneously from the comfort of his or her office or bedroom.
Will non-medical users need training to use it effectively?
Training? Not quite. This wont be different from recording blood pressure or blood sugar on the go on the mobile phone. Better still, just like recording a Whatsapp message on the phone and sharing it. If these other activities required any prior training, it is all in a matter of minutes.
Developing Al in an environment with unstable electricity and limited equipment sounds daunting.
How did you navigate it? What resources did you lack the most?
Indeed, very daunting and to be honest, more challenges lie ahead but they are not unsurmountable. The most important thing is to set your goals, define your targets, thereafter, be determined and stay focused as the obstacles, which I call distractions come, including human distractions. The most important thing is to assemble the brain, acquire sophisticated computers, hire external clouds for data storage and manipulations, which sometime would be in Dollars. You can always overcome non-steady power supply through solar panels and inverters. I think the most annoying thing is the frequent damage to your computing units through terrible and unpredictable power surges and of course frequent theft cases which keeps setting you back. But alas, slow and steady win the race.
All said, I still believe the hardest task is in assembling the brains and keeping them together. Young skilful individuals are highly mobile.
Funding has been a major obstacle. What aspects of the project were delayed or compromised due to insufficient support? How much funding would be required to fully scale this innovation?
Yes, funding has been a major obstacle but our experience is that if you build integrity and you have a convincing proposal, funders will eventually smile at you at some point. Many people would like to support ingenious ideas, particularly with prospect of commercialisation. Funding slowed things down but didn’t stop us outrightly. I remember that following the preliminary data that we had using TETFUND seed grant, our second attempt at sourcing for funding saw us putting up a grant proposal to the tune of about a Billion Naira to the National Institute of Health of the United States. The grant proposal was accepted for processing but on technical ground, we crashed out. The same work we have now done without taking a Billion Naira from any body but by looking inward, meaning that our achievement now is worth over a Billion Naira.
Do you believe Africa has been scientifically misrepresented in global medical datasets? What consequences does that have for African patients?
Certainly yes . We have always said that there is inequity in health data generation and health research execution. The greatest burden of diseases is in the middle and low income countries, Nigeria inclusive but most of the funding in health research is taking place in the developed world such as Europe and America. Africa, carrying a significant burden of diseases receives less than five percent of the total expenditure on global health research, a figure that is disproportionately low. We must set aside a significant part of our GDP to fund research into our local problems, medical inclusive.
Given that users will record ECGs on mobile phones, what data-protection measures are built into your system?
Confidentiality, privacy and safety of data for future retrieval are key to what we are doing. Reasonable measures have been taken to guarantee this and also protect from hackers and malware attacks.
Are you complying with any international health data standards?
Oh yes. First and foremost, the research work has been approved by the National Health Research Ethics Committee of the Federal Ministry of Health of the Federal republic of Nigeria. Similarly, research and ethics committee of all the participating University Teaching Hospitals also reviewed and approved the research work before commencement. By this I mean, the scientific, moral and ethical standards were examined before approval was granted. I was also privileged to obtain a certificate on international health standards from Harvard University years back where we were trained on cross border international health data transfer. Further more, we complied with international ethics on Al data analysis.
Should African governments require that diagnostic tools be validated for African physiology before being deployed?
What steps can policymakers take immediately?
Oh yes, that’s what Standards Organization of Nigeria and NAFDAC are established to achieve. Then, more recently, local content policy of the federal government of Nigeria. Every new medical device should be assessed to meet our local needs and peculiarities before being thrown into the market for public consumption. Where possible, we should also insist on technology transfer in the development of such devices, just like America is doing with China and India with China. Nigeria cannot be a dumping ground for all technologies for people to exploit our growing and consumptive population.
Have any Nigerian or international tech firms approached you for collaboration?
Sure, we have received several requests for partnership and collaboration over the years of the project. Some partnership can be very stifling and counterproductive. So, you really have to shine your eyes very well to know which ones to embrace and the ones to reject along the line. We are still open to such as the offers keep coming but what is important is for us to define our own minimum and irreducible conditions that we want met in order not to blind our own long term vision and agendum for the project.
What kind of partnership would accelerate deployment the most?
At this point now, capital investment so that we can go the whole hog, even more so that we are set to leverage on what we already know to deploy similar AI application to solve other health-related problems and non-health-related societal problems, including social, commercial, educational, religious, political, tourism and even entertainment industry.
How many lives do you estimate could be saved yearly if Africans had access to accurate ECG interpretation powered by AI?
Worldwide, millions of ECGs are recorded yearly. In Nigeria, hundreds of thousand of ECGs are recorded annually. This is mostly in urban areas and are usually necessary as a result of the presenting complaints of the patients. Ours should increase access since there are mobile telephones and networks even in the rural areas. It would therefore be possible to record ECGs in the rural areas and use an App on the mobile phone to interpret, in addition to relaying the ECG signal through mobile network to the central ECG Laboratory.
It is not going to be only mere accurate ECG interpretation, but accurate cardiovascular disease diagnosis, even in the remotest parts of the world, like heart failure and heart attack. Accurate ECG diagnosis is not an end in itself, but a means to an end amongst several other diagnostic tools. We now live in a globalised world where Africans now reside and work in different parts of the world and this AI design would also be very useful in evaluating their sicknesses as they live in other parts of the world without Cardiologists of African extraction or background.
Could this reduce the number of sudden cardiac deaths often reported in the region?
Certainly yes. This is because someone with sudden, excruciating chest pain can now quickly run an ECG on himself in the comfort of his car, office, or home and get an instant AI-generated report on whether the chest pain is due to any of the causes of sudden cardiac death, such as heart attack, heart failure, or heart block, and then seek immediate intervention at a nearby competent hospital.
Do you foresee this innovation reducing the healthcare gap between urban and rural communities? What infrastructure is needed to achieve that?
This innovation would certainly improve access to health care, particularly cardiovascular care. People can now have an idea about their cardiovascular health without necessarily seeing a cardiologist just at the tip of their fingers. The innovation is bound to bridge the gap in access between urban and rural areas, since up to half of the population in rural areas now have a mobile phone, including access to an internet network.
Looking 10 years ahead, how do you think this single innovation might reshape cardiology practice in Africa? What role will AI play in the next generation of diagnosis?
Ten years ahead, I see faster, more efficient and accurate cardiovascular care. I see individual patients or citizens playing more central role in the diagnosis and treatment of their cardiovascular conditions. I see lesser need for human minder or carer at homes for individual patients since there will be better automated measures to remind patients about medication timing and exercise routine.
If the Nigerian government gave you full backing today, what breakthrough would you pursue next? Would you expand into other areas of African-specific diagnostics?
Oh yes. What is important is to have digital data for different clinical scenarios. We can repeat the same feat in eye, surgical, diabetic, stroke, hypertension and obstetric care and so on. We have the template which can even be extended beyond medicine.
What message do you have for young Nigerian or African scientists who believe impactful research can only be done abroad?
No research is useful if it does not make any impact on the way we live and make society better. Impactful research relevant to a community is best done by researchers or institutions closest to that community. Therefore, African problems are best understood by Africans, and we are most suited to proffer solutions. Some research requires cutting-edge technology, such as space research, nuclear research, and so on, but many others require only simple scientific inquisitiveness and sound methodology to arrive at universally acceptable conclusions. Even where everything cannot be done here in Nigeria, there is room for collaboration across borders and continents, which is exactly what we have done over the last two decades on world comparative normal ECG, and which dovetailed into the present AI-propelled diagnostic algorithm. To put it succinctly, impactful research is being done and can be done from Nigeria and any other country for that matter. It only requires a painstaking and methodical team of researchers following a well-thought-out protocol.
