Tracking the patterns

Dominique Brodbeck and Markus Degen from the Institute of Medical and Analytical Technology already have experience on research projects outside the natural sciences. For example, they developed software for hospital managers to trace patient movements between examinations. The data is used to plan future infrastructure and help improve daily routines in both the clinic and administration. The current project also focuses on optimising hospital operation. Together with a software developer, the two researchers have created a computer program that shows the energy consumption of imaging examinations in hospitals. The focus is on radiological departments, whose MRI and CT devices consume a million kilowatt hours per year – the equivalent of around 250 households. Brodbeck speaks of two reasons for the “GreenRad” project, run in cooperation with Basel University Hospital: “One is economic: with ever more equipment and rising prices, electricity bills in radiology are soaring. The other is an ethical drive to reduce CO2 footprints and with them, dam- age to the environment.”

However in order to cut the bills, electricity consumption is not the only factor. The maximum power capacity must also be paid for and that depends on how much total power the hospital uses at one time. “In order to reduce energy consumption, you have to be able to say how much energy is used by one machine per time unit, per examination and in standby,” explains Brodbeck. “It is not enough to just look at the current curve. That is why we use different data sources.” In addition to electricity consumption, the scientists also collect operation data from the cooling system, activity records for each machine, as well as clinical data on the type of examinations. GreenRad can correlate and present these data quickly and interactively, allowing users to monitor energy consumption patterns and to determine their frequency and characteristics. IT engineer Degen sees this approach as fundamentally different compared to conventional handling of large amounts of data: “Data mining is usually used with correlation algorithms to find patterns automatically. We use humans as a recognition algorithm; people not computers recognize recurring processes and patterns. Our thesis is that there are many questions where only people can make sound decisions. We present the data on a plate and provide the software tools.” In order for the data to end up on that plate however, a considerable effort is required, says Degen: “The data isn’t consistent at all at first, either in time or in content: it must be aligned.” Once this has been done, the data must be visualized in such a way that the relevant information can be extracted. Brodbeck describes the requirements of the program: “It must be interactive and visually clear so that you can find information quickly if you have an idea.”

The system test in Basel already revealed energy consumption trends: in standby mode, a machine consumes constant current, just like a TV. During an examination, power consumption rises to peak level from less than a second up to a few minutes. Here, Degen sees improvement potential: “A high-level control system could synchronize all machine operations of less than one second so that high current consumption occurs in series and not in parallel.” This could significantly reduce the maximum power capacity needed. GreenRad’s key ability to make patterns visible comes to the fore when reducing total electricity consumption and thus the CO2 footprint. Degen cites an example: “We counted up to twenty calibration sequences during CT scans. This could probably be reduced.” He is sure that the program has even wider potential: the researchers used their software to test the log files of computer programs. Because they wrote the program, they can determine all variables and data types themselves. As a result, there are virtually no limits to the possible applications.