To some people visualisation is just making some pretty pictures to go with a grant proposal or press release about your research. In some quarters it does have more to do with art than science but, hey, we need to get the public interested in science somehow.
Another aspect of visualisation is finding something in the huge torrent of data that emerges from today’s big experiments and system level simulations. The field combines data mining, statistics, signal processing and any other kind of mathematics that you can find to reduce the data and identify features that tell us something. Only then can the pretty picture play a part in communicating it. After all, it is believed that about half the reasoning ability of the human brain is devoted to processing visual stimulation.
Visualisation teams or specialists used to be thought of as a service or support department within a research facility. But Raghu Machiraju, associate professor of computer science and engineering at Ohio State University, has come out of the support world to get down and dirty with the application specialists to do the actual science. He sees visualisation as an instrument to observe the natural world and then to make representations so that others can understand what he has seen.
Starting with fluid mechanics, he has developed important collaborative relationships across disciplines – and now, like many other scientists, has found some of the most interesting challenges in the biomedical field. His work has got him noticed and he is already regarded as an extremely important contributor both to the applications field and visualisations generally. His profile will grow even more this year when he is co-chair of the IEEE Visualisation Conference. Colleagues predict that his star is rising and that a lot more important work is still to come.
Professor Chris Johnson, director of the Scientific Computing and Imaging Institute at the University of Utah, says Machiraju had already made a considerable impression on the visualisation community and there was every sign that he would continue to make important contributions.
He says: ‘Raghu is well known in the visualisation field, particularly for his work in flow visualisation and vortex detection. More recently he has been expanding into biomedical applications and has been doing some interesting work on trying to find classified features within biomedical images. The biologists have recently found new techniques with very high resolution imaging systems; they are coming to computer scientists to help analyse these images, so we are seeing collaborative teams working together.
‘As we have gotten larger amounts of data, and that has become more complex with higher dimensions, we need new visual abstractions to take this data and filter it so we can understand it better. Before great discoveries in science there is often the creation of a new tool and visualisation is one of those new tools.
‘One of the things that people see about Raghu is his amazing curiosity and intensity. When he is trying to solve a problem he comes at it every which way, and he has many tools himself. He is very sophisticated mathematically, he knows a lot about computer graphics and computer science, and he also goes and learns the applications: the fluid dynamics or the biology. He is really looking at problems from a holistic point of view, which is not very common.
‘He may not sound very ambitious, but that is part of his humble and unassuming nature. In terms of the intensity of his curiosity and the work that he puts into his science, he is very ambitious. This is echoed in terms of solving research problems and making a contribution and he has already done so. He has all the tools needed to make a real impact and I think he is already a way along that path.
‘People often look at things through the tools that they know very well, but one of the things that make Raghu so good is that he looks at things from many different ways at once and that is very powerful.
‘He may not know as much about, say, biology, as the top experts in that field, but he learns a considerable amount of it. The best visualisations are done when we are working closely in an effective collaborative team.
‘Raghu is a very friendly person; he is one of the nicest people you could meet. He has very charismatic personality and he laughs a lot. He likes really spicy food and sometimes he will bring his own hot sauce to a meal.’
The hot sauce is a good clue. Machiraju was born in the city of Vijawada in Andhra Pradesh, south India (the Indian state reputed to have the spiciest cuisine in the country). His father is an economist who studied in the US, before returning to India to teach and work for the Central Bank of India. His mother was a homemaker. Education was very important in the family and all of his siblings have advanced degrees. Machiraju studied hard, particularly at mathematics.
He says: ‘It was a big thing in India in those days that you had to be good at maths to succeed. Now it is very different. I did well and I was not even 15 when I got admitted to the Indian Institute of Technology (IIT), but I didn’t have a very good personal experience so I went back home.’
He is very modest about this point, but about 300,000 apply for the entrance examination to the IIT every year, and only 5,500 get places. He says: ‘I guess it is hard to get into, but you can get there if you have the right formula. My age was not that unusual, because the educational impetus in India is so much. I guess I could not take the rigours of staying away from family, I was just too young. I ended up going to the University of Delhi a few years later where I studied electrical engineering, mostly because I enjoyed physics and thought there was more interesting mathematics in electrical engineering. There was a lot of peer pressure in India in those days to study engineering and there were not as many options as there are in the US. We went into engineering, because that is where the jobs are.’
After graduating from Delhi he returned to Bangalore to study at the Indian Institute of Sciences (IIS), the country’s leading centre for postgraduate science education. He decided to study computer science. Computers were starting to become more common and at the IIS he had access to the latest generation of machines, so it seemed like an interesting field into which to move.
After completing his masters he decided to try for a PhD programme at a US University, as his father had done before him, and ended up studying mathematics at the University of Tennessee.
He says: ‘I was there for about six months and I didn’t like it. Then a job came up at Control Data Corporation (CDC) and I applied for it. I worked there for two years as an applications programmer. CDC was a well-respected pioneer of high performance computing (Seymour Cray was its chief designer in the 1960s and 70s), but by the late 1980s it was in decline. His bosses at CDC suggested he look for some other job as cuts were looming, so he moved to the Ohio Supercomputer Center. He did many different things here, from writing applications to creating an operating system for transputers. Eventually, though, he became bored. He decided that it was time to do a PhD, which he did at Ohio State while working as an engineer in the supercomputing centre. It was here that he first got involved with visualisation. He had his first flurry of published papers before gaining his PhD in 1996 with a thesis about volume rendering and picked up an award for research excellence.
He says: ‘I got into visualisation because I liked the guy who taught the course, Roni Yagel. He eventually left academia to go into industry, but he was a nice guy to work with and I liked the kind of problems he had in mind, so I ended up joining his group.’
The next stage in his career was to get a faculty job and he took up a position at Mississippi State University in the NSF Centre for Computational Fluid Simulation.
He says: ‘This place had a very good reputation for work in fluid dynamics and I started working closely with the fluid dynamics applications people on visualisation. Indeed, I still do.’
After three years at Mississippi he returned to a faculty position at Ohio State and started to broaden the scope of his research work into biology.
He says: ‘Here I work closely with people in radiology and cell biology. Most of my work is taking their data and trying to extract features. I use a lot of techniques, such as machine learning and all kinds of algorithms, to try to understand what is in the data; it’s a lot of maths.
‘Visualisation is not just making pictures; it’s about getting ready to make pictures. There is so much stuff in the large data sets and you have to fish it out. It’s not just algorithms in a vacuum; you need to have some notion about how the data was generated and what the underlying equations are. We have had some results in fluid dynamics and we have worked a lot on vortex detection algorithms. I worked with David Thompson, who is an aerospace engineer at Mississippi State, and the combination of an engineer and a computer scientist working on the problems has resulted in us finding some new ways to think about the problems.
‘Now I am working on the ability to reconstruct three-dimensional structures at a cellular level. I support a lot of phenotyping studies. Biologists are interested in how a particular phenotyping experiment will change the structure of a particular organ, so I try to make three-dimensional constructions of those organs so they can see the differences. My focus really is turning towards biology these days, but I still do some fluid dynamics.’
Machiraju says that over the years the role of the visualisation specialist has changed considerably. Many research teams would see visualisation as a service and just deposit their data with the visualisation team for them to make some nice pictures. But he believes the pure visualisation team is going out of fashion. One of the reasons is, of course, cost. The huge teams, with their own management structures, that existed at big government laboratories are now shadows of their former selves and some of what they used to do is being transferred to a specialist within the research team. He says the modern approach is to say: ‘We have some data, can you help us out?’
He says: ‘Since the mid-90s, really, the amount of data being produced has been ballooning and people wanted to have some understanding of what was actually going on. But there is the feeling in some parts that visualisation is a luxury: if you have it, great, if you don’t then, oh well. Part of it is that visualisation has tried, but it has not yet reached its full potential. There have been many successes, but also we have not cracked the big one, how you make sense out of tons and tons of information. I think it needs some change in how we educate our young researchers; we need to get more into understanding what the data is all about, and how you work with the data. Also, we need to reassure people about what they see on the screen and how to represent uncertainty on the screen. Many experienced researchers have noticed that there is a gap in expectations and there has been a flurry of work in addressing some of these issues, such as how do you train new researchers who go into areas like data compression and do they fully understand it.
‘There are always new ways of representing things that are more intuitive and are more appealing to the cognition, but at the same time you need to pack more meat into the image and make it more meaningful. Any comparisons have been made with the great artists and how they pack so much more into the representation of a scene.
‘But for some of us, it’s more about the way the data was created, the underlying equations and the statistics. There are three or four camps in the field that people have created and hopefully, over the next decade, we will be able to produce people who have a full understanding.
‘You need a general good education that allows you to understand a field to a sufficient level, which means understanding the mathematics and how the display works, and also being adroit enough to jump into a new field and learn about it. It is these people who will go beyond the issue of producing display screens and they will go forward into the problem-solving aspect.’
He says that one of the issues is trying to generate smart data by concentrating the most detailed analysis on the areas more likely to produce interesting data. He said: ‘It’s about being clever about the data and producing less of it. All fields have the same problems of producing too much data, so you have to incorporate more knowledge into the data. I have had some excellent collaborators and, once you have their trust, you are seen as their partner and they share their data with you, rather than outsourcing the visualisation to you. But it’s the hardest thing to do to get to know people. The problem with science is that you don’t always know what it is you are looking for half the time.’
One of Machraju’s ambitions is to create a strong signal processing framework for visualisation, and he has received NSF funding for a collaborative project around this. Another of his ambitions is to find a way of incorporating variation into the representation of biological structures, because in the real world there are a lot of variations possible within any given organ or organism. In fact, he has many more ambitions, because there are so many challenges out there that he wants to overcome.
He says: ‘I am more of an engineer than an artist, but really I am just an applied mathematician. I like to break things down in a reductionist way and put them back together. I am always more intrigued by the real world around me, rather than the pure abstract.
‘What I am trying to do is to try and explain and help people understand. I think of visualisation as a computational microscope, trying to get down to the way things are and understand the structure. I am trying to understand what is inside and then to present it. I think there are some beautiful structures there of which I can make images. That is why I am having so much fun with cellular biology, because there are so many things you can look at; there are so many different kinds of cells and you have to think of ways to represent them.
‘Most of my work has been done at systems level with mice models and I am looking at arrangements of cells rather than inside cells, which is important to things like the mechanisms of cancer progression. What I try to do is extract three-dimensional structures and annotate them. Many times there are multispectral inputs.’
Machiraju says he has been approached to work at other institutions, but has so far turned them down. He likes the lifestyle in Columbus, Ohio, and is keen to raise his family in a cosmopolitan place. He is happy with the way the university has treated him and there are lots of opportunities to collaborate. He hopes to be made a full professor soon, but it does not seem to be the main thing occupying his mind at present.
Playing host to the IEEE Visualisation Week in Columbus, and the chance to meet new people doing interesting things with visualisation, are things that he is much more passionate about at the moment.
Education
1982 Delhi University, Delhi, India B.S, Electrical Engineering.
1984 Indian Institute of Science, Bangalore, India MS, Automation
1996 The Ohio State University, Columbus, Ohio PhD, Computer and Information Science.
Employment
1986-87 Control Data Corporation, Applications Analyst
1987-91 The Ohio Supercomputer Center, Senior Engineer
1991-96 The Ohio State University, Graduate Research Assistant
1996-99 Mississippi State University, Assistant Professor
2000-03 The Ohio State University, Assistant Professor
2003 to date The Ohio State University, Associate Professor
2004-05 Mitsubishi Electric Research Laboratory, Consultant