Lionel Tarassenko FREng - An Engineering Biography

Professor Lionel Tarassenko FREng

Moving into mobiles

So, after his graduation, Lionel went to work for Racal. Radio communications and engineering, the company’s business, had appealed to him when he was at school. He joined Racal at a time when mobile telephony was a new idea. This was before Racal set up Racal-Vodafone, which turned into Vodafone and went on to become the world's biggest mobile telephone business.

On joining Racal, Lionel embarked upon the usual graduate project required of new recruits. Though it was supposed to be finished in eight weeks, his project caught on and lasted for a couple of years! It helped to set the foundations for Racal’s move into digital signal processing – and Lionel then worked on Racal’s first speech coder.

The work got Lionel hooked on a subject that has occupied much of his subsequent research career. “I really fell in love with signal processing,” he explains. It wasn't just the number crunching that appealed, but the fact that in order to make speech processing work, you have to know something about speech. “You have to understand how speech is produced. So you have to model the vocal tract.”

When the call came to do his year’s military service back in France, Lionel had more time to think about his new interest in signal processing and human physiology. The year gave him thinking time, not to mention the chance to discover a love for teaching.

A born educator

This turn of events is one of many chance occurrences that Lionel credits with changing his direction. Whilst posted to a radar station on a French Air Force base, Lionel took the opportunity not just to teach fellow conscripts, but also to read up on medical electronics.

Influenced by a tutor who used ultrasound to measure blood flows, Lionel had already become interested in biomedical electronics as an undergraduate. Still a minority pursuit in the early 1980s, when Lionel wanted to do a PhD in the area the only opening was in the Department of Paediatrics at Oxford University.

“I must be unique for a Fellow of The Royal Academy of Engineering, because my doctorate, strictly speaking, is in the Department of Paediatrics,” says Lionel. He was guided by an enthusiastic Professor of Paediatrics, Peter Tizard, who had an interest in biomedical instrumentation.

Scanning the brain

Lionel picked up a young scientist prize for a conference paper about his research – on the use of electrical impedance tomography to image the brains of premature babies. However, in the 1980s medics weren't so enamoured of medical technology and Lionel didn't fancy becoming a somewhat superior technician, keeping monitors working in a hospital. So he turned to mainstream electrical engineering in Oxford's Department of Engineering Science.

This did not mean forgetting what he knew about the brain. Neural networks were enjoying a renaissance as a research topic at the time. While much of the attention was on using neural networks to study how the brain might process signals, Lionel anticipated other possibilities: “what I saw was a new model for a form of signal processing that would be based on learning about the signals in order to process them in a different way. We started to use neural networks for signal processing.”

So Lionel had moved on to conventional engineering challenges, but he was using tools with a physiological inspiration. Neural networks are trained in a way which mimics how our brains work.“You extract meaningful information from the data, in the learning process, without explicitly having to write down a set of rules. So you learn to recognise faces, you learn to recognise features in speech, as you learn to recognise certain patterns in jet engines or in biomedical data.”

Right time, right place

Lionel’s doctorate in brain monitoring and work on speech processing led him to work on neural networks and signal processing. It was also a happy coincidence that neural networks became a hot topic at the right stage in his career. “I was privileged enough to be elected to a chair in electrical engineering at the age of 39, because I happened to have done research in an area which all of a sudden was exploding.”

Another coincidence came when Sharp, the Japanese electronics company, set up its European research headquarters in Oxford. The company asked some local academics for ideas. Lionel, who had already talked about neural networks to one of the people who went on to work for Sharp, wrote a report on the subject.

Coincidence piled on coincidence when Sharp read a competitor’s catalogue of new products and saw a reference to a microwave oven with a neural network. “We want one of those,” said Sharp, and asked Lionel to come up with something. The result was LogiCook, the first microwave oven that you didn’t have to tell what kind of food you wanted to cook. (The ‘rival’ turned out to be no more than a glorified toaster.) The neural network recognised the ‘signatures’ of various foods, and the oven turned itself off without user intervention when the food reached the right level of “doneness,” as Sharp described it. Thus one of the company’s most successful products was born and the first to be designed outside Japan. Sharp also ended up recruiting two of Lionel’s best students.

Looking for differences

While neural networks are of interest to researchers in themselves, it is what you can do with them that matters. This is where Lionel got interested in another application of neural networks, novelty detection. This is not so much the pursuit of style gurus, as an attempt to study signals for indicators of things that really matter, the tell-tale signs of something out of the ordinary.

Most systems generate sensor signals that are highly predictable. Who cares if a jet engine, or a patient’s heart, is throbbing away regularly hour after hour? You only really want to know when something new is happening that might presage trouble ahead. “In biomedical engineering, for example, one of the things that you want to detect is abnormal physiology.” So why not train your system on normal behaviour? It can then, in effect, ignore that and look for occurrences of abnormality.

This work started to draw Lionel away from microwave ovens back into biomedical engineering. He credits Professor Sir Michael Brady with this, when they started discussing ideas on how to detect tumours in mammograms. The approach they took was to extract parameters from different regions in a mammogram and to learn what constitutes ‘normal values.’ The system could then detect and flag up novel areas so that the diagnosticians could pay attention to a particular aspect of the mammogram.

Out of the ordinary

This introduces an important issue for Lionel: “I have never tried to design a fully-automated patient-monitoring system. It is always decision support, smart prompting. Someone who reviews mammograms will see hundreds in a week. A novelty detection system will simply say ‘hey, before you go onto the next one make sure you have a look at that and decide whether it is normal tissue or possibly a malignant tumour’.”

Why not fully automate this process? “Because I knew as an engineer I could design a system that would be 99.9% accurate, but the cost of getting that 0.1% wrong is such that I still can't design automated systems that are fully automated. You still need the human in the loop to make the final decision.”

It was when describing his work on mammography that serendipity stepped in again. Someone from Rolls-Royce approached Lionel after a talk and asked about using the same approach to look at X-ray images of turbine blades. That idea came to nothing but they discovered that there might be better applications of neural networks – in ‘condition monitoring,’ checking the health of jet engines while they are flying. So Lionel found himself monitoring the health of patients and jet engines.

Soon after, two seemingly academic activities turned into new businesses. Lionel went back to Isis Innovation, Oxford’s successful spin-out promoter, which had already approached him for business ideas. The result was Oxford BioSignals which now has two divisions, one for biomedical engineering applications and another to handle the work on jet engines, with hundreds of laptop PCs now used in Rolls-Royce’s engine testing facilities around the world.

Mobilising medicine

Along the way, Lionel also set up a business that grew out of an invitation from his old colleagues at Racal, now with Vodafone. Get in touch, they said, when you have ideas for medical uses for mobile phones. Lionel did just that as soon as the stars lined up and 3G telephone networks could carry enough data. He wanted to see if mobile phones could help people to manage such chronic diseases as diabetes. The Vodafone Group Foundation funded clinical trials of a project that also involved the Division of Public Health at the university and the Oxford Centre for Diabetes, Endocrinology and Metabolism.

Would patients be prepared to buy into a service that helped them to control their illness and reduce the risk of going into hospital or of developing the long-term complications of diabetes, such as blindness and lower-limb amputations? Lionel set up a new company, now called t+ Medical that would thrive or die depending on the answer to this question.

Initial success

The company may have won an award for its diabetes management system – software that allows a patient to collect and transmit blood glucose data over a mobile phone for immediate processing and feedback – but that is no guarantee of commercial success. It turns out, though, that the healthcare world is interested in using the technology for community nurses to monitor patients and concentrate on those in need of attention. In this way, the NHS can manage more patients far more cheaply than through nurses’ visits or hospitalisation.

“The model has changed to a service provision enabled by a technology, as opposed to consumer healthcare,” explains Lionel. “I didn't know that six months ago.” The lesson from yet another unexpected finding? You have to be nimble, he says.

The combination of firstclass research and business achievement won Lionel the Silver Medal from The Royal Academy of Engineering last year. The medal “recognises an outstanding personal contribution to British engineering that has led to market exploitation by an engineer who is under the age of 50”.

Biomedical engineering

While Lionel clearly remains passionate about research, teaching and, in his spare time, bringing ideas to market, he still finds time to check out what’s going on in ‘brain science.’ Indeed, in the middle of all this work on his ‘day job’ he spent 18 months helping to coordinate the physical science side of the Foresight Cognitive Systems Project, a major effort that brought together researchers in information technology and the brain sciences and resulted in a book that charts the state of play in this fascinating area of interdisciplinary research.

Lionel’s interest in uniting engineering and biomedicine isn’t just out of a belief that biomedical engineering is a wonderful way of working across disciplinary boundaries to come up with new ideas. He also sees the broader benefits. It is a great way of persuading young people and the wider public that engineering is relevant to them. “Biomedical engineering touches everybody’s life.” For example, he asks, “why do people survive severe traffic accidents? Because patient monitors or brain scanners tell the medics what is going on inside the body.”

Lionel might try to claim that it is another coincidence that puts him in such a strong position as biomedical engineering rushes into the mainstream, with initiatives like an Institute of Biomedical Engineering at Oxford for example. This time, though, it is more of a case of reaping the rewards of spending more than a quarter of a century working hard to persuade engineers, and medics, to take the subject more seriously.

A change of view

These days, says Lionel, “biomedical engineering is really considered to be a very important discipline by the medics themselves. That wasn't the case 25 years ago.” Indeed, back then, he adds, “we were seen by the medics as glorified technicians.” Today, Oxford’s medics are, he says, extremely supportive of plans to make biomedical engineering much more central to Oxford’s research.

Some might say ‘about time too’. Lionel is more charitable. He realises how long things take to change, especially at universities like Oxford. After all, as he points out, he is only the university’s second Professor of Electrical Engineering. Engineering itself even took time to catch on at Oxford – although the Department of Engineering Science celebrates its centenary next year. With multi-disciplinary endeavours such as the Institute of Biomedical Engineering at the centre of its plans, engineering at Oxford is in very good shape as it contemplates the next hundred years.

Achievements

Born 1957 in Paris. Graduated in 1978 with a BA in Engineering Science from the University of Oxford. Joined Racal in 1978 and worked on speech coding and feasibility study for cellular radio. Returned to the University of Oxford in 1981 and was awarded doctorate in Medical Electronics in 1985. Young Scientist Prize for paper on brain imaging in pre-term infants in 1984. Appointed in 1988 as University Lecturer and Tutorial Fellow, St. Hugh’s College, University of Oxford. 1996 Elected a Fellow of the IEE. 1996 British Computer Society Medal for neural network analysis of sleep disorders. 1996 IEE Mather Premium for work on neural networks. 1997 Elected to the Chair of Electrical Engineering and Professorial Fellow, St. John’s College, University of Oxford. 1999 Founder director of ThirdPhase Ltd. 2000 Founder director of Oxford BioSignals Ltd. 2000 Elected a Fellow of the Royal Academy of Engineering. 2001 Rolls-Royce Chairman's Team Award for Technical Innovation. 2002 Founder director of e-San Ltd (now t+ Medical). 2002 Scientific Co-ordinator, Foresight Cognitive Systems Project (Office of Science & Technology, DTI). 2002 Deputy Head of Department, Department of Engineering Science, University of Oxford. 2005 UK E-Health Innovation Award for best use of e-health to empower patients. 2006 Institute of Engineering & Technology IT Award for Data Fusion System for Early Detection of Patient Deterioration. 2006 Silver Medal, Royal Academy of Engineering.

Biography - Michael Kenward OBE

Michael Kenward has been a freelance writer since 1990 and is a member of the Ingenia Editorial Board. He is Editor-at-Large of the Science|Business online magazine. Prior to this, he worked on the New Scientist for 20 years and was editor of the magazine throughout the 1980s. He also worked with Lionel on the Foresight Cognitive Systems Project.