Care and Cure

Visual Coding at TU/e

During the past 50 years an explosive growth spurt within electrical engineering has radically altered our lives. Society has become dependent on the products produced by electrical engineering – in fact, no other discipline has had such an impact. The next 50 years will see even more radical changes in areas such as communications, healthcare, energy supply, mobility and production automation.

In view of these developments, the Department of Electrical Engineering has identified three areas of research with high societal relevance:

  • The Connected World (Developments in Communication)
  • Smart and Sustainable Society (Working Toward a Sustainable Society)
  • Care and Cure (Electrical Engineering in the Healthcare Sector)

The department’s employees and students have achieved resounding successes in close collaboration with industry and social partners in each of these research fields. Furthermore, with Brainport Eindhoven - a leading technology hotspot in Europe - close by, TU/e  is clearly the place to be for academic researchers who ant to change our world for the better.

We talk to three key employees in the Visual Coding & Architectures group to find out more about TU/e. Prof.dr.ir. Peter de With, the head of the department, PhD candidate Francesca Manni and Assistant Professor Fons van der Sommen.

​​​​​​​WHAT MAKES TU/E SPECIAL?

Peter: Before my professorship at TU/e, I worked for Philips Research and also held a full professorship at the University of Mannheim in Germany. Close lines of communication with the high tech industry on your doorstep, broad changes for funding and the opportunity to escape the relative anonymity of a large university were what brought me back to Eindhoven. That close connection to industry is absolutely unique.

Francesca: I couldn’t agree more. After an internship with Philips Research in Eindhoven, I briefly returned to Italy. But the universities in Italy are not close to industry. So I’m pleased to be back in Eindhoven and working on unique hyperspectral image analysis projects with a medical focus. We are part of a European project called ASTONISH (Advance Smart Optical Imaging and Sensing for Health) with many industrial partners.

Fons: I don’t have any experience with working abroad yet, but my research on computer-aided detection and diagnosis of early oesophageal cancer would not be possible without the close relationships we have with industrial partners and the large hospitals in the Netherlands. Working together with medical doctors and showing them how the technologies we are developing will help them to work more efficiently is deeply rewarding.

image

TU/e clearly offers research staff challenging assignments. What exciting projects have you worked on during the past 12 months?

Peter: Healthcare, surveillance and automotive are all areas where our intelligent camera technology has contributed. However, I’m particularly proud of our detection system for IEDs (improvised explosive devices), i.e. concealed roadside bombs. The Dutch Ministry of Defence asked us to  develop a system for detecting these explosive devices as they account for most of the deaths among our soldiers in countries like Iraq, Afghanistan and Mali. We started working on this project about three years ago and have now reached the stage where we can detect suspicious objects of a diameter of 10 cm at a distance of 40 to 50 metres. The Ministry compared our detection system to 5 similar projects that are being developed in other NATO countries and judged it the most effective. It is also the only solution that operates in real time. We are now prototyping a device for the European Defence Agency. This system, which is the result of a collaboration with ViNotion and CycloMedia, is going to save many lives.

Fons: I am hugely excited by our recent breakthrough in Volumetric Laser Endomicroscopy (VLE) – a world first! This is an imaging technology that allows you to literally look inside human tissue at a highly accurate resolution (measured in micrometres). Oesophageal cancer is our current area of focus. Our research has proved for the first time that computers are much better at visually scanning granular grey-tone images and recognising small differences - early indicators of cancer - than human doctors who are experts in the field. In tests, expert medical specialists achieved a success rate of 77%. Our computer program performed outstandingly and achieved an 92% detection rate. The technique will help doctors detect cancer at a very early stage when treatment is both cheaper and far more effective. The program, which has been validated by multiple experts, analyses all the data at a breath-taking speed of up to 30 images a second. The computer sees every single pixel and every single image. It then uses an algorithm to decide whether action needs to be taken. Deep learning will improve the system’s performance as we gain more experience. This breakthrough would not have been possible without the partnership between TU/e, large hospitals and an industrial producer of endoscopic equipment.

Francesca: We have just started our second year of research into tissue tracking for use by surgeons when treating cancer. We started from a blank sheet of paper last year and have already reached the stage where we can experiment with actual patients. This new tool looks beyond the visual spectrum and allows us to see more than the surgeon during an operation. I’m working closely with the Dutch Cancer Research Institute and it’s really exciting.

image

Working on research projects like these must be deeply satisfying. What developments do you see in three years’ time?

Peter: Living field labs and deep learning are areas where we can make real progress. Once you have established these platforms, you can accelerate innovation enormously. And we have a pressing need for good academics to help us achieve our objectives.

Fons: Proper validation of the systems we develop is one of the challenges. Another one is deciding what operating points you build into your system. You don’t want to reduce sensitivity to a level where cancer cases slip through, but you also don’t want to worry people unnecessarily. Increasing our understanding of deep learning will help here.

Francesca: I will be delighted if we can produce a realtime imaging tool for cancer detection during surgery. And develop novel imaging techniques for detecting cancer in patients in a non-invasive manner. The funding is there and new cameras are on the way, so I’m very optimistic.

TU/e’s contribution in these areas puts it in a class of its own. Would you recommend TU/e to other academics looking for a research position?

Peter: Of course. With all the projects we have on the go, we need a steady supply of good people. Thanks to the links with industry and the availability of many cooperative funding cases, our researchers have every opportunity of engaging in ground-breaking research and publishing papers on unique work in reputable journals. Eindhoven is also a great place to live.

Fons: I would echo that. Another thing I’d say is that the University believes in developing its people. I have been offered a custom-made tenure track which will allow me to engage in research at foreign universities. That will be a big help in furthering my career.

Francesca: TU/e has opened up a world of opportunity to me. As a young researcher, life couldn’t be better. I’m working on really important projects, the funding is there, the links with industry are fascinating and there are all kinds of opportunities for developing my career. Eindhoven’s buzz and nightlife is also very attractive for people of my age. TU/e is an ideal place for people who want to engage in research of real societal value.

What unresolved question or dream inspires you?

Peter: My dream is detecting cancer at the earliest possible stage. The sooner we detect cancer, the greater the chance of survival.

Fons: Having pilot and demonstration systems running in hospitals in 5 to 10 years and refining machine learning to achieve 100% accuracy. Machines can analyse the tiniest details and scan images in a fraction of the time required by human specialists. This allows these systems to learn from millions of examples and they will provide medical specialists with additional information, enabling them to make better informed decisions. 

Francesca: We need enough data to validate the tool and make our research more efficient. Surgeons are willing to  cooperate but the time available to them for collecting data is limited. You have to publish to convince them of the value of your work.

image