Virtual Scientific Laboratories Debuting on the Internet

For more than a decade, scientists have traveled to Berkeley to use the unique high voltage electron microscope at Berkeley Lab. Soon, the Laboratory will take the facility to the user.

No, the three-story-tall microscope (the most powerful in the nation) has not been put on wheels. Rather, the Laboratory has created a set of interactive, online computing tools that will allow scientists to manipulate the instrument, conduct experiments, and view images from their own offices, via the Internet.

In-situ electron microscopy-inducing and observing changes to a sample in the microscope-had never been done remotely until a demonstration by Berkeley Lab scientists at the annual international meeting of the Microscopy Society of America in Kansas City. Two thousand miles from Berkeley, researchers used a computer to take control of the microscope, heat an advanced alloy specimen, and observe the ensuing progression of structural changes on the computer monitor.

The demonstration heralds a new phenomenon, the advent of online "virtual laboratories." As computer and network technology progresses, major scientific resources like the Berkeley microscope will become accessible from all over the world. Collaborators at different locations will conduct an experiment and discuss results while it is in progress.

The team bringing the electron microscope online is led by computer scientist Bahram Parvin and Michael O'Keefe, the deputy head of the Lab's National Center for Electron Microscopy. Other team members include John Taylor, Brian Crowley, Doug Owen, Ken Westmacott, Bill Johnston, and Uli Dahmen.

Remote control of machines, in and of itself, is not new. Remote operation of an electron microscope, however, accomplishes something thought to be almost impossible. That's because of the difficulty of overcoming the slight but inevitable delays in time that occur with Internet use.

During in-situ microscopy, things are moving fast. Down at the atomic and molecular level, matter drifts and changes shape. Furthermore, the microscope is equipped with an experimental chamber for subjecting a sample to a range of different conditions such as heating, cooling, tensile strain, and varying atmospheric pressures.

Right now, says Parvin, an operator sitting next to the microscope has to constantly adjust dials just to keep the area of the specimen under study within the field of view. Focusing requires continuous effort. Imagine trying to do this over the Internet, which has a time lag analogous to the voice lag that occurs during an international phone call. By the time the remote operator's commands reach the microscope, they are too late.

To sidestep this limitation of computer networks, Parvin's team is automating onsite the positioning and focusing of the microscope through the development of advanced computer vision algorithms.

Explains Parvin, "You start with what to a computer is an indiscriminate field. You then detect and lock onto objects of interest. This is computer vision. Very soon, from a remote location, computer vision will self-calibrate the microscope, autofocus it, and compensate for thermal drift. Underlying this is a complex package of algorithms dealing with shape analysis, background measurements, wavelet transform, and servo-loop control.

Experimenters at remote locations will be able to "drive" the microscope. They will be able to change magnification, scan the sample, alter its orientation, and trigger a range of experimental conditions. Collaborators will do this through a computing environment that includes the necessary video-conferencing tools. Parvin also is developing another set of computer vision tools with Berkeley Lab life scientists. These not only detect objects of interest-for example, single DNA molecules-but lock onto and track them.

As a starting point for this technology, Parvin's team has authored algorithms for a range of shapes. The team also has created a tool for drawing an outline around uniquely shaped objects and tracking them.

To see a demonstration of computer vision tracking of DNA, check out this short 167K mpg movie.