Computer Scientists Aid Analysis of Biological Images

Posted: 

By Vanessa Spates

raghu_tractography_hires.jpg
Interdisciplinary work is a lot more complicated than one might think, but it’s enlightening and immersive, says Raghu Machiraju, a computer science and engineering associate professor who is helping researchers at The Ohio State University Medical Center learn more about basic mechanisms of disease, including cancer, and improve patient outcomes.

He first worked with Gustavo Leone, of the Center for Solid Tumor Biology and Ohio State’s Comprehensive Cancer Center - Arthur G. James Cancer Hospital and Richard J. Solove Research Institute, and Michael Ostrowski, in molecular and cellular biochemistry, to learn the basic mechanisms underlying tumor initiation and progress and with Michael Knopp in radiology to delineate brain structures. Machiraju found ways to improve the processing and analysis of acquired images, such as MRI and microscopy results.

“Through imaging, statistical analysis and machine learning, we actually showed that many of the cell types in the microenvironment surrounding a mammary duct undergo changes in morphology in addition to changes in the DNA. This work required us to create statistical models of nuclear shape in both normal and mutant microenvironments. We especially focused on the changes that are likely to occur during tumor initiation. This subtle change is hard to detect. We continue to verify our findings using transgenic mice that fluoresce in a confocal or multi-photon microscope,” Machiraju says.

The researchers are deploying similar methods in work with Chandan Sen and Sashwati Roy at the Dorothy M. Davis Heart and Lung Research Institute to study basic mechanisms underlying the endothelium, cells that line blood vessels, in the formation of those vessels in normal and oxygen-deprived wounds, which are slow to heal. They also work with Clark Anderson from the Heart and Lung Research Institute to better understand the dynamics of the immunological protein transport in the placenta. They use imaging and machine learning to study an assembly of cells and different patterns.

Machiraju and his collaborators also have developed methods to extract white matter, or neural tissue, connectivity from MR images of the brain.

“Knowledge of connectivity facilitates superior surgical planning and treatment of neurological diseases,” he explains.

Biology, Machiraju says, is reductionist in nature and very different from computer science, where abstraction and intuition are central.

“Biology is certainly an observational science and is very much hypothesis driven. The latter aspect is very foreign to many computing researchers, who build tools to explore and interact with data,” he says.

Despite the major differences between the two sciences, the team has found ways to work together to learn more about the basic molecular mechanisms of cancer and to help students understand the biological structures and the instrumentation in addition to the implications of the research.

“All of my students know how to operate scanners and microscopes; know the subtleties of the processes (staining tissues for fluorescence); and are adept at reading leading biology, radiological and neuro-scientific journals,” Machiraju says, explaining that he has former students who are fellows at Harvard Medical School, MIT/Harvard Broad Institute and NIH. “All of them are really well-trained in an interdisciplinary fashion. They are well-versed in computer science, statistics and their domain of interest in biology.”
Vanessa Spates is an undergraduate student in journalism.

 

Category: Research