Science & Research

Research Spotlight: March 10, 2016

News Editor
Thursday, March 10, 2016

Computer vision technology aids in classification of leaves

Though the vein structures of leaves may help identify plants, botanists seldom refer to them due to the difficulty involved in deciphering their variations. But in a study released this month, researchers found that computer vision and artificial intelligence technology can assist botanists in classifying leaves according to their vein structures.

The paper, co-authored by Thomas Serre, assistant professor of cognitive, linguistic and psychological sciences, was released in the Proceedings of the National Academy of Sciences.

Computers learned to discern certain types of leaves by shape and vein pattern based on an initial set of 7,600 digital images of plants that had been chemically enhanced to emphasize their veins.

The software easily distinguished among relevant patterns — the algorithms sorted new leaves into their taxonomic family with 70 percent accuracy and into their order with roughly 60 percent accuracy. In a University press release, the study’s lead author, Peter Wilf of Pennsylvania State University, said that the software’s ability to identify family and order is “an incredible achievement.”

Algorithm tracks how cells move

Researchers who study cancer and autoimmune disorders may benefit from a new technique developed by a team of University scientists who found a way to track cell movements within the body.

In the past, scientists hoping to study cell movement used a technique called traction force microscopy. In this method, as cells move through an artificial space — a 2D surface or a 3D gel — researchers capture images that they can then use to calculate any forces the cells generate as they move. Still, for researchers to conduct calculations, they must know mechanical qualities such as the stiffness of the artificial environment.

Now, through a process called “mean deformation,” scientists can make these calculations without having to know the mechanical properties of an artificial environment, thus allowing them to test the movement of cells on surfaces that better mimic the human body, said Christian Franck, assistant professor of engineering, in a University press release.

Franck said that he and his colleagues plan to release the software that powers the technology for free online, so that other researchers may use the technique by uploading their own 3D images.

IVF cited for high twin birth rate but could also reduce it

In vitro fertilization has often been credited with the record-high rate of twin births in the United States. But earlier this month, Eli Adashi, a professor of medical science at the Alpert Medical School, argued that IVF could play a role in lowering the rate to more natural levels.

Twin births may come with complications such as preterm births and low birth weight, which is why many physicians consider IVF’s artificial contribution to the rising rate of twin births problematic. The U.S. Centers for Disease Control and Prevention reported a record high of 33.9 twin births per 1,000 live births in 2014.

With the steep cost associated with IVF, mothers and doctors aspire to have the invasive method of fertility assistance work on the first try. Doctors implant more than one embryo into a mother, with the hope that if one dies, others will survive.

This, Adashi said, is where the solution lies. To reduce twin births, he argues, states should advocate insurance coverage of IVF transfers of only a single egg.

“Combining a direct path to IVF with the judicious use of superovulation/IUI should go a long way toward curtailing the national twin birth rate,” Adashi wrote in his editorial, which was published in the March volume of the American Journal of Obstetrics and Gynecology.