Teaching Computers to See and Understand Disease

On this issue of the IBM Patent Leadership Series, we are featuring Maria Gabrani, a member of the IBM Research team based in Switzerland.

The IBM Patent Leadership series marks IBM’s 25th consecutive year leading in patent grants. The series consists of innovation stories from remarkable IBMers who help make this company what it is and has been for over a century: a collection of inquisitive, restless, determined humans who drive the culture of innovation within and beyond our walls.

At the age of four, Maria Gabrani caught pneumonia and had to spend a month in the hospital. “Albeit a difficult experience, it sparked a passion in me to cure people and solve medical problems.” Healthy and back on her feet at home, she would chase her little brother around the house until he allowed her to perform medical tests and exams on him.

Maria turned her passion for medicine and love of mathematics into a thriving career. These days, she uses image processing, pattern recognition and machine learning techniques to solve medical problems in areas ranging from computational pathology to drug discovery.

Maria’s current research focuses on extracting meaningful information from images of biopsies and of structures removed through surgery. She holds more than 25 US patents related to making sense of images and diagrams.

Patent US9053532, for example, is a method for evaluating automatically an image from a scanning electron microscope (SEM), and has been applied to virus and disease identification, and even to the testing of medicines and vaccinations. “The advancement of image processing, pattern recognition and machine learning to model and solve medical problems is an iterative process that can significantly improve our understanding of disease development and treatment effectiveness. I am convinced that medicine needs mathematical and computational power to reach its curative potential.”

Medical images are by far the largest and fastest-growing source of data in the healthcare industry. In fact, it is estimated that they make up at least 90 percent of all medical data, a figure overwhelming to even the most data-drenched specialists. (And, perhaps, even to the pathologists tasked with assessing these images manually.)

Maria wants to ease the workload of these pathologists by teaching computers to see and understand medical images. “By combining deep learning and dictionary learning technology, not only can information be extracted from stained tissue images, but differences and similarities of cells and cell formations can also be quantified in a structured way.”

This technology enables the study of differences in human tissue that make it difficult to personalize medicine. “We’re at ease when software recognizes where we’re located and gives us directions to our destination. Just imagine if a similar concept was applied to tissue images and other medical data—for heart disease or cancer—to help find the best diagnosis, prognosis and treatment for each individual.”

In celebration of Women’s Month, we asked Maria to share her thoughts on women’s progress:

“Women’s progress comes when we do not need to have special discussions or measures to make any decisions; when we discuss about qualifications, requirements, responsibilities with respect to work, without considering gender but with sensitivity to context of life in general; when we discuss about general life requirements and responsibilities without considering gender and sacrifices with respect to work; when we can celebrate the positive qualities of a human without any gender or other connotations. Equality comes in all directions.

“To use technical terms, to have stable optimal it has to be a global one. Where we currently stand depends a lot on the demographics and the situation. So I would not generalize. In the current way of changing things: from realization of need, to awareness, to action, legislation and cultural shift to status quo, we can at least proudly claim that the journey has started and in some cases it is advanced.”

Maria Gabrani is a Research Staff Member for the IBM Research team in IBM Switzerland. She has led several international, interdisciplinary projects, across IBM and with external partners, leading to significant contributions to IBM’s next generation products. Maria Gabrani has received several awards including two Outstanding Technical Achievement awards, from IBM Corporation, a Technical Achievement Award, from IBM Research Division and several best paper awards, among them from NASA/Goddard Space Flight Center, Image Registration Workshop.

Find out more about the foundation of IBM’s Patent Leadership.

If you are interested in finding out more about opportunities at IBM, follow us on Facebook, Twitter, LinkedIn, Youtube and Instagram


Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.