Big data being used to identify prostate cancers, best treatments

Some prostate cancer patients unnecessarily undergo surgery or harsh treatments because science fails to identify the differences between slow-growing and aggressive forms of the disease.

Researchers at Case Western Reserve University and Johns Hopkins Medical Institutes are developing technology that allows patients to safely choose to do nothing, opt for relatively mild treatments or take drastic measures.

The project is one of several Anant Madabhushi, an associate professor of biomedical engineering at Case Western Reserve and director of the Center for Computational Imaging and Personalised Diagnostics, is involved in.

Madabhushi brought the centre and a dozen researchers to Case Western Reserve from Rutgers University this academic year. The center's overall mission is to develop the new science of using big data to identify different forms of diseases and the most effective treatments.

''The data does exist,'' Madabhushi said. ''If someone has a disease, we can detect it. The issue is what to do next. We're detecting half-a-million cases of breast and prostate cancers each year. How do you treat them? What are the options? What is the right treatment?''

Madabhushi and Robert Veltri, an associate professor of urology at Johns Hopkins, have won a three-year, $1.2 million Department of Defense grant to develop a guide to steer prostate cancer patients to the most effective treatment.

The medical community estimates 241,740 prostate cancer cases were diagnosed nationally last year, and the number is expected to grow annually as the population ages and detection improves.

The researchers are concentrating on identifying the 15 percent of prostate cancers considered very low-risk. Sufferers are likely to survive 20 years, depending on the treatment they receive.

''We don't want to expose someone to a therapy that won't work for him,'' Madabhushi said.

Not only may patients suffer unnecessarily, the costs can be astronomical. For example, radiation therapy always destroys healthy tissue. Rectal toxicity can be caused by radiation spillover from prostate treatment, he explained. Use of a proton gun to treat prostate cancer reaches about $125,000 per patient.

In their search for cues that help identify very low-risk prostate cancer and the best treatments for each form, the researchers are looking at data from 300 men, half who chose immediate surgery after diagnosis and half who chose to have their cancer monitored.

Researchers at Johns Hopkins are focusing on molecular biomarkers from tissue samples of prostate cancers. They are looking at a set of five unique biomarkers to separate the very low-risk group from the low-risk, intermediate-risk and high-risk groups.

The Case Western Reserve researchers are undertaking a quantitative computerised image analysis-based study of the organisation and structure of diseased tissues.

Specifically, they are looking for differences in the shape and size and architecture of nuclei within the tissue, as well as the shape of the cross-sections of glands in the samples. All can indicate the disease's aggressiveness and characteristics.

The data comes from digitised images of biopsy slides. A simple biopsy image of prostate cancer digitised via a whole slide scanner can be several gigabytes in size. In a typical biopsy procedure, 10 to 12 samples are removed and, after those are laid out on slides and digitized, that translates to hundreds of gigabytes of data.

''All of this data is a challenge and opportunity; the words go hand in hand,'' Madabhushi said. ''How we deal with it all is a challenge, but the opportunity is in what it can tell us. There's so much we don't know and there's only so much a pathologist can tell us. There's a limit to how much time a pathologist can spend analyzing a glass slide.''

The researchers write and employ algorithms, which are programs encoding complex mathematical equations to model complex structures, to explore the digitized images.

They've developed some of the algorithms needed to analyze prostate cancer and expect to fine-tune them with this project. They will develop a new algorithm that combines the molecular biomarker measurements and the computer-extracted morphological measurements.

''The beauty of this technology to predict the aggressiveness of cancer is we don't need to take new tissues that would be destroyed, but we are working only from digitized images and we can go back to look at them repeatedly,'' Madabhushi said.

''This could work anywhere in the world, including developing countries where molecular testing is expensive. This can be easily done,'' he said.

Madabhushi's group is doing similar work on the many forms of breast cancer and is exploring use of the technology on colon cancers with researchers at the Case Western Reserve School of Medicine.

Within the Madabhushi lab, Mirabela Rusu, a research associate, has been awarded a Department of Defense post-doctoral fellowship grant aimed at identifying the hallmarks of prostate tumors that are resistant to radiation or recur after treatment.

Satish Viswanath, a research assistant professor of biomedical engineering, has been recommended for funding for a Department of Defense post-doctoral fellowship grant to combine data gathered from mechanical imaging and magnetic resonance imaging, to improve diagnosis and prognosis of prostate cancer.

Each fellowship provides $125,000 for research over two years.