This article is taken from at Guelph,the University of Guelph online service for further dissemination.The article may be accessed at the following link:
Computers Can Help Identify Diseased Cells
Prof’s algorithm analyses cellular characteristics
A crucial step in diagnosing many diseases and cancers is the examination of suspected lesions in medical images by a radiologist or cells under a microscope by a pathologist. The problem is that even skilled physicians can make errors, sometimes diagnosing a disease or cancer that doesn’t exist, or missing one that is present – and both kinds of mistakes can have serious outcomes for patients.
But change is on the way. Biomedical engineering professor April Khademi is conducting research that makes big strides towards reducing those errors and providing consistently accurate diagnoses using algorithms in radiology and pathology images.
Until recently, pathologists used a microscope to look at cells, but new wholeslide scanners generate very high resolution images that allow them to see much more detail. “You can see the cells and the nuclei on the computer screen enlarged 20,000 times,” says Khademi, who joined U of G in January.
However, all that detail can be difficult to manage and evaluate – and that’s where Khademi’s work comes in. She has developed algorithms (mathematical formulas) that allow the computer to analyze the data contained in the image. It might, for example, assess the roundness of the cells’ nuclei, or count the number of cancerous cells in a particular area. “Algorithms can quantify the information in an objective way,” she explains. “The process will give the same answer every time.”
She adds, “This is a brand new field. I am trying to create mathematical tools that mimic human perception, but in a better way. For example, a pathologist might look at a piece of tissue and say it has a rough texture. The algorithm can determine that the tissue is 80-per-cent rough, or can assess the image of a cell nucleus and determine that it has a 0.4 amount of roundness where the pathologist might just be able to say it was not very round.” These calculations can provide a foundation for better diagnoses.
Khademi, who was born in Saskatchewan but grew up in Toronto, says her engineer father fostered a love of math in his daughter and sons, who are also engineers. “I could do logarithms when I was in Grade 1,” says Khademi.
She earned her bachelor’s and master’s degrees in electrical engineering at Ryerson University, where she also met her future husband. She was awarded the Governor General’s Gold Medal for her research on computer-aided diagnosis of mammograms.
While completing her PhD at the University of Toronto, Khademi was awarded an NSERC Canada Graduate Student D3 grant (the highest level). Her studies there, done in collaboration with staff at Sunnybrook Hospital, developed algorithms to analyze the amount of white-matter lesions in a person’s brain.
“Almost everybody has some white-matter lesions in their brains,” she says. “These are areas where the tissue is not actually dead, but it is being starved of oxygen. We know that they can be a precursor to strokes and to dementia, but we don’t know much about how or why.” Her algorithms quantify the number and size of the lesions, providing objective data that can then be followed up over time to see which people have strokes or develop dementia. Those results may provide guidance for future treatment and more personalized medicine.
About 50,000 Canadians suffer new or recurrent strokes each year, which means on average a stroke occurs every 10 minutes in Canada. Stroke is the third highest cause of death behind heart disease and cancer, costing the Canadian economy roughly $3.6 billion a year in physician services, hospital costs, lost wages and decreased productivity.
Khademi hopes her research will lead to the development of new technological innovations that help reduce mortality rates, long-term disability and the economic burden associated with stroke. She is presently continuing this research and applying her algorithms to a large image database collected from patients across Canada.
During her PhD studies, she received additional awards: one was the L’Oreal-UNESCO Award for Women in Science. “They flew me to Montreal for a big gala event,” Khademi recalls. But she was even more impressed by how she was treated when she twice won a Google Canada Anita Borg award. Those awards earned her a week in New York City and another week in San Francisco, where she had the opportunity to take part in presentations and be wooed by Google.
Two days after graduation, Khademi started working as an algorithm development specialist for GE Healthcare at the Pathology Innovation Centre of Excellence. Her work not only involved developing the algorithms that give meaning to digital images, but educating pathologists and others about the new technology.
Two years later, Khademi was hired by Pathcore, Inc., a Toronto-based digital pathology software company, where she focused on designing innovative products and algorithms as a senior scientist and product manager.
She was offered a position at U of G soon after. “I am honoured to continue my research in medical imaging technologies that I hope will transform the way medicine is practiced and improve the lives of patients with stroke and cancer,” she explains.
When she’s not working on ways to enhance medical technology, Khademi loves hot yoga and in-line skating; she also plays the flute. “I’m also very involved in volunteer work to promote math, science and engineering to girls and young women,” she says. “Did you know that only 10 per cent of the students in electrical engineering are women? I want to help change that.”