Coding the Cure for Cancer

In this special guest feature, Emily Walsh, the Community Outreach Director for the Mesothelioma Cancer Alliance, discusses how artificial intelligence is helping to detect cancers as well as the role AI is playing in the processing and sorting of data collected by cancer researchers. The Mesothelioma Cancer Alliance works to raise awareness about mesothelioma and connect patients with the resources they need. Emily connects the alliance with online communities to bring information on mesothelioma and asbestos to the world. Her advocacy efforts center around the dangers that asbestos still poses today to help prevent future outbreaks of mesothelioma cancer.

How AI, computers, and big data are on the forefront of cancer detection, diagnosis, and treatment.

Human innovation and excellence can only go so far, but where the prowess of human intelligence stops, machine learning continues. Since Richard Nixon’s National Cancer Act of 1971, doctors and scientists have been racing for a definitive cure for these diseases, with little luck in finding a foolproof cure.

Nearly 35 years later, The Cancer Genome Atlas (CGA) project was established, creating a concerted push to map the human genome against cancer data to find patterns in people with similar cancers, and use that data to propel the oncology field into a curative state, rather than a defensive one.

This half century war on cancer has yielded incredible amounts of information, certainly more than any team of humans could ever process and understand, let alone draw conclusions and cures from. Too much data should never be a bad thing, and data scientists suspected that somewhere among the multitude of scans, patient info, genetic markers, and medical records lay useful information.

Data Organization

The National Cancer Institute (NCI) launched a competition in 2013 to design a cloud-based system to organize their data from the CGA. In 2014 the NCI publicized their three contract winners; the Broad Institute in partnership with UC Berkeley and Santa Cruz received $7 million, the Institute for Systems Biology in partnership with Google was awarded $6.5 million, and Cambridge-based bioinformatics company Seven Bridges Genomics received $6 million.

Even technology company Intel started their own cloud system called the Collaborative Cancer Cloud after losing their bid for the NCI contract. These systems make the encrypted and organized CGA data available to research teams nationwide, allowing for further developments in ways to process these heaps of information.

Though these attempts to create a usable platform for all of this information certainly take the first step at turning CGA data into cures for cancer, other research teams around the world started with AI, before getting to data organization.

AI + Cancer

Artificial Intelligence poses a huge potential for the cancer world. Because of a machine’s innate attention to minutiae and the ability to process more information in short amounts of time than a human could in a lifetime, they are valuable assets to cancer fighting teams.

The Gustave Roussy Cancer Treatment system is one of these such designs that results in faster tissue analysis, and therefore decreases the time from cancer appearance to diagnosis. Using AI systems to meticulously analyze radiological scans for abnormalities results in better attention to detail, and alerts radiologists to suspect areas on their scans based on previous data.

These AI systems are initially fed cancerous scans from previous patients and are programmed to distinguish between diseased and non-diseased cells. Then, they’re given new scans and tasked to determine if someone has cancer on their own.

In some instances, these computers equipped with AI are accurate 95 percent of the time, compared to 65 percent achieved by humans. In fast-moving and aggressive cancers, this could mean the difference between life or death.

Mesothelioma, an aggressive form of internal cancer, moves slowly and then progresses very quickly. From asbestos exposure, the only known cause of mesothelioma, to a cancer diagnosis, the time frame can be as long as 50 years. After diagnosis, a mesothelioma patient’s life expectancy is usually less than two years.

Scottish firm Canon Medical Research recently received a £140,000 grant to develop an AI system specifically to detect and diagnose malignant pleural mesothelioma before humans can, to prolong patient’ life expectancies. Concentrated applications of AI to specific diseases seem to be the highest performing systems, as opposed to broader machines.

Closing Thoughts

Humans may have finally designed a way to turn the tide against cancer, but these advances don’t come without drawbacks. Radiologists are worried that advanced machinery may result in job loss, and some skeptics don’t think AI will live up to its advertised potential.

Even IBM’s Watson, one of the first large AI endeavors, can’t seem to crack the treatment phase of cancer care, and has repeatedly made incorrect, dangerous suggestions in tests.

However, AI isn’t meant to be an autonomous operator when it comes to disease care. Dr. Robert Schier, RadNet radiologist, suggests thinking of AI as “an army of fellowship-trained radiologists with photographic memories, IQs of 500 and no need for food or sleep,” rather than a cure-all replacement.

Sign up for the free insideBIGDATA newsletter.