Monthly Archives: September 2016

Depth sensor to approximate

MIT researchers have developed a biomedical imaging system that could ultimately replace a $100,000 piece of a lab equipment with components that cost just hundreds of dollars.

The system uses a technique called fluorescence lifetime imaging, which has applications in DNA sequencing and cancer diagnosis, among other things. So the new work could have implications for both biological research and clinical practice.

“The theme of our work is to take the electronic and optical precision of this big expensive microscope and replace it with sophistication in mathematical modeling,” says Ayush Bhandari, a graduate student at the MIT Media Lab and one of the system’s developers. “We show that you can use something in consumer imaging, like the Microsoft Kinect, to do bioimaging in much the same way that the microscope is doing.”

The MIT researchers reported the new work in the Nov. 20 issue of the journal Optica. Bhandari is the first author on the paper, and he’s joined by associate professor of media arts and sciences Ramesh Raskar and Christopher Barsi, a former research scientist in Raskar’s group who now teaches physics at the Commonwealth School in Boston.

Fluorescence lifetime imaging, as its name implies, depends on fluorescence, or the tendency of materials known as fluorophores to absorb light and then re-emit it a short time later. For a given fluorophore, interactions with other chemicals will shorten the interval between the absorption and emission of light in a predictable way. Measuring that interval — the “lifetime” of the fluorescence — in a biological sample treated with a fluorescent dye can reveal information about the sample’s chemical composition.

In traditional fluorescence lifetime imaging, the imaging system emits a burst of light, much of which is absorbed by the sample, and then measures how long it takes for returning light particles, or photons, to strike an array of detectors. To make the measurement as precise as possible, the light bursts are extremely short.

The fluorescence lifetimes pertinent to biomedical imaging are in the nanosecond range. So traditional fluorescence lifetime imaging uses light bursts that last just picoseconds, or thousandths of nanoseconds.

Blunt instrument

Off-the-shelf depth sensors like the Kinect, however, use light bursts that last tens of nanoseconds. That’s fine for their intended purpose: gauging objects’ depth by measuring the time it takes light to reflect off of them and return to the sensor. But it would appear to be too coarse-grained for fluorescence lifetime imaging.

The Media Lab researchers, however, extract additional information from the light signal by subjecting it to a Fourier transform. The Fourier transform is a technique for breaking signals — optical, electrical, or acoustical — into their constituent frequencies. A given signal, no matter how irregular, can be represented as the weighted sum of signals at many different frequencies, each of them perfectly regular.

The Media Lab researchers represent the optical signal returning from the sample as the sum of 50 different frequencies. Some of those frequencies are higher than that of the signal itself, which is how they are able to recover information about fluorescence lifetimes shorter than the duration of the emitted burst of light.

For each of those 50 frequencies, the researchers measure the difference in phase between the emitted signal and the returning signal. If an electromagnetic wave can be thought of as a regular up-and-down squiggle, phase is the degree of alignment between the troughs and crests of one wave and those of another. In fluorescence imaging, phase shift also carries information about the fluorescence lifetime.

The communication connections established by the top 500 Android apps

MIT researchers have found that much of the data transferred to and from the 500 most popular free applications for Google Android cellphones make little or no difference to the user’s experience.

Of those “covert” communications, roughly half appear to be initiated by standard Android analytics packages, which report statistics on usage patterns and program performance and are intended to help developers improve applications.

“The interesting part is that the other 50 percent cannot be attributed to analytics,” says Julia Rubin, a postdoc in MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), who led the new study. “There might be a very good reason for this covert communication. We are not trying to say that it has to be eliminated. We’re just saying the user needs to be informed.”

The researchers reported their findings last week at the IEEE/ACM International Conference on Automated Software Engineering. Joining Rubin on the paper are Martin Rinard, a professor of computer science and engineering at MIT; Michael Gordon, who received his PhD in electrical engineering and computer science in 2010 and remained at CSAIL as a researcher until last July; and Nguyen Nguyen of the engineering firm UWin Software.

Different operations performed by the same mobile app may require outside communication, and rather than try to coordinate shared access to a single communication channel, the app will typically open a separate communication channel for each operation.

The researchers analyzed the number of communication channels opened by the 500 most popular mobile apps and found that roughly 50 percent of them appear to have no bearing on the user experience. That doesn’t necessarily translate directly to the quantity of data exchanged over those channels, but for short sessions of application use, the portion of transmitted data irrelevant to user experience is also as much as 50 percent.

Across longer sessions, in which large files are transferred to the phone — by, say, music- or video-streaming services — the percentage of data transmitted covertly steadily diminishes. But covert communication channels remain open.

Piercing the veil

Mobile applications are usually proprietary: Their source code is not publicly available, and their developers frequently take pains to disguise the details of the programs’ execution, a technique known as obfuscation.

Algorithm speeds up complex modeling from days to hours

To work with computational models is to work in a world of unknowns: Models that simulate complex physical processes — from Earth’s changing climate to the performance of hypersonic combustion engines — are staggeringly complex, sometimes incorporating hundreds of parameters, each of which describes a piece of the larger process.

Parameters are often question marks within their models, their contributions to the whole largely unknown. To estimate the value of each unknown parameter requires plugging in hundreds, if not thousands, of values, and running the model each time to narrow in on an accurate value — a computation that can take days, and sometimes weeks.

Now MIT researchers have developed a new algorithm that vastly reduces the computation of virtually any computational model. The algorithm may be thought of as a shrinking bull’s-eye that, over several runs of a model, and in combination with some relevant data points, incrementally narrows in on its target: a probability distribution of values for each unknown parameter.

With this method, the researchers were able to arrive at the same answer as a classic computational approaches, but 200 times faster.

Youssef Marzouk, an associate professor of aeronautics and astronautics, says the algorithm is versatile enough to apply to a wide range of computationally intensive problems.

“We’re somewhat flexible about the particular application,” Marzouk says. “These models exist in a vast array of fields, from engineering and geophysics to subsurface modeling, very often with unknown parameters. We want to treat the model as a black box and say, ‘Can we accelerate this process in some way?’ That’s what our algorithm does.”

Marzouk and his colleagues — recent PhD graduate Patrick Conrad, Natesh Pillai from Harvard University, and Aaron Smith from the University of Ottawa — have published their findings this week in the Journal of the American Statistical Association.

Modeling “Monopoly”

In working with complicated models involving multiple unknown parameters, computer scientists typically employ a technique called Markov chain Monte Carlo (MCMC) analysis — a statistical sampling method that is often explained in the context of the board game “Monopoly.”

To plan out a monopoly, you want to know which properties players land on most often — essentially, an unknown parameter. Each space on the board has a probability of being landed on, determined by the rules of the game, the positions of each player, and the roll of two dice. To determine the probability distribution on the board — the range of chances each space has of being landed on — you could roll the die hundreds of times.

If you roll the die enough times, you can get a pretty good idea of where players will most likely land. This, essentially, is how an MCMC analysis works: by running a model over and over, with different inputs, to determine a probability distribution for one unknown parameter. For more complicated models involving multiple unknowns, the same method could take days to weeks to compute an answer.

Shrinking bull’s-eye

With their new algorithm, Marzouk and his colleagues aim to significantly speed up the conventional sampling process.

“What our algorithm does is short-circuits this model and puts in an approximate model,” Marzouk explains. “It may be orders of magnitude cheaper to evaluate.”

The algorithm can be applied to any complex model to quickly determine the probability distribution, or the most likely values, for an unknown parameter. Like the MCMC analysis, the algorithm runs a given model with various inputs — though sparingly, as this process can be quite time-consuming. To speed the process up, the algorithm also uses relevant data to help narrow in on approximate values for unknown parameters.

In the context of “Monopoly,” imagine that the board is essentially a three-dimensional terrain, with each space represented as a peak or valley. The higher a space’s peak, the higher the probability that space is a popular landing spot. To figure out the exact contours of the board — the probability distribution — the algorithm rolls the die at each turn and alternates between using the computationally expensive model and the approximation. With each roll of the die, the algorithm refers back to the relevant data and any previous evaluations of the model that have been collected.