A Digital Breakthrough for Cancer Diagnosis

Digital signal-processing techniques extract information from DNA to identify changes that occur as cancer develops

7 December 2009


tech Photo: Jess Molina
IEEE Fellow K.J. Ray Liu with a screen from his ensemble-dependence model program.

Medical checkups often involve screening blood to measure the levels of such components as glucose, cholesterol, and triglycerides. One day an ordinary screening test may also include checking your DNA to tell whether you might be developing cancer.

That’s the vision of IEEE Fellow K.J. Ray Liu and his team at the University of Maryland, in College Park. They are using digital signal-processing techniques to extract information from DNA to identify changes that occur as cancer develops, which they hope will ultimately lead to the ability to predict whether cells will become cancerous.

“Nowadays a doctor can tell you, for example, what your cholesterol level is and express it by a number,” Liu says. “Hopefully, through our work, one day a doctor will be able to give you a number related to cancer—whether the number is within a normal range, whether the number shows cells are transitioning to the cancerous stage and preventative treatment is needed, or whether the number is high and you need to watch for cancer developing in, say, your liver or breast.”

Liu has written or cowritten 10 books and more than 500 research papers. He  coedited the Handbook on Array Processing and Sensor Networks (Wiley–IEEE Press, 2009).

“There is a paradigm shift in cancer diagnosis under way, from a completely biological process to engineering a digital world,” Liu said in March when he described his work at an IEEE media event in New York City. “We are seeing the dawn of a digital revolution in cancer diagnosis at the gene and protein level by using digital signal processing.”

Cancer is a leading cause of death worldwide and the number of cases is increasing, according to the World Health Organization. WHO reports that the number of cancer deaths around the world is projected to jump from about 7.9 million annually to almost 12 million by 2030. New cases of cancer are expected to rise from 11.3 million per year to 15.5 million in 2030.

As the disease develops, cancerous cells release unique proteins and other molecules that can serve as early indicators. Such biomarkers display alterations of patterns at the cellular, molecular, or genetic level. Correctly identifying protein biomarkers for cancer holds enormous potential for early detection, diagnosis, and treatment—which is particularly important for cancers of the skin, breast, cervix, mouth, larynx, colon, and rectum.

Thanks to the sequencing of the complete human genome, there have been many advances during the past decade to help identify biomarkers. One technique involves the application of microarrays that can measure and translate the expression level of thousands of genes simultaneously. Those expressions can be processed into digital signals, Liu points out.

Liu’s microarray technology translates seemingly random biological information in DNA into an expression of data that can be read by a computer. His “ensemble-dependence model” looks at the microarray DNA data or proteins via a mass spectrograph by classifying them into different clusters; analyzes the dependence between genes and proteins; and assesses their behavior and interaction. Each cluster contains specific genes that have a well-defined relationship to one another.

That dependence is used in Liu’s model to classify normal and cancerous samples. The system is then applied to identify cancer biomarkers using several real cancer data sets, including prostate and gastric cancers. Liu’s group used its model-driven approach to uncover the relationship between the global gene expression profile and a subject’s health. That could lead to the ability to predict cancer development in the lungs, stomach, colon, prostate, and ovaries with an accuracy rate as high as 98 percent, he says.

“We are trying to use very simple equations in very simple models to explain the very complex phenomena of cancer development,” Liu says. “We’re developing different models to try to classify whether there is cancer, and if so, at what stage.”

Unlike the two-year time frame it typically takes to conduct R&D for a new high-tech product like an iPod, a much longer and more expensive process is required for medical research, Liu notes. He estimates it will take 10 to 20 years to see the application of his early-prediction technique, which he began developing in 2002.

He says he hopes that in the next 5 to 10 years “digital testing for cancer can supplement the traditional biological testing to offer a reliable second opinion, improve cancer detection accuracy rates, and reduce false alarms.”

FOR MORE INFORMATION You can watch Liu’s talk at the media briefing.

IEEE membership offers a wide range of benefits and opportunities for those who share a common interest in technology. If you are not already a member, consider joining IEEE and becoming part of a worldwide network of more than 400,000 students and professionals.

Learn More