Machine learning could help revolutionize cancer diagnosis


Data driven medicine used by IBM’s Watson and several other start-ups like InvitroCue are revolutionizing cancer diagnosis.

Credit: Pixabay
Credit: Pixabay

Machine learning is a subfield of computer science, that grew out of the quest for artificial intelligence. It is so pervasive in today’s world that you probably use it often in daily life, without even realising it. Machine learning has given us self-driving cars, effective web search, recommendations that you get when you visit web sites or social media sites, face detection in a digital photo album, stock trading etc.

Machine learning enables computers to analyze vast amounts of data and automatically detect patterns and features, or make predictions regarding certain conditions. In a dynamic disease like cancer, gauging and diagnosing such a complex heterogeneity is the biggest challenge. After decades of cancer research, it has become increasingly clear that no two patients’ cancers are exactly the same, and even within one person’s tumor there is a wild diversity of cells. Accurate and quicker diagnosis is very crucial in rapidly progressing cancers. And with the vast data of research and clinical data that has been compiled over the years, who can make sense of this deluge of data? It would take eons for a human, but minutes for a supercomputer.

Looks like Watson helps crack mysteries not only for Sherlock!

A woman in Japan, initially diagnosed of acute myeloid leukemia, a rare form of blood cancer, underwent chemotherapy. However, her recovery after the treatment was unusually slow, leading her doctors in University of Tokyo’s Institute of Medical Science to suspect of a different form of leukaemia. In a mysterious medical case, that looked straight out of a Dr. House series, conventional tests failed to show any sign of it. Dr. Tojo’s team, turned to IBM’s Watson, which proved to be a life saving move. Watson, a cloud based AI-powered computer system,  developed by IBM, cross referenced this case against 20 million oncological records, and concluded in a mere 10 minutes that the patient suffered from a different type of leukaemia than originally diagnosed, following which they started a new treatment regimen, which was far more successful.

“We would have arrived at the same conclusion by manually going through the data, but Watson’s speed is crucial in the treatment of leukaemia, which progresses rapidly and can cause complications,” said Dr.Tojo.

IBM join hands with CognitiveCare in China to customize cancer therapy

21 hospitals across China have adopted Watson for Oncology, in an initiative to provide customized, evidence-based approach to treatment due to the unique nature of the disease. Watson for Oncology not only brings the world class oncology expertise of Memorial Sloan Kettering Cancer Center, but is also a continually updated learning system, that will only get better over time.

“Healthcare in China is transforming at a rapid pace but the world’s most populous country faces numerous challenges as it struggles to cope with a precipitous rise in cancer and other diseases,” said Nancy Fabozzi, Principal Analyst of Transformational Health at Frost & Sullivan. “Optimum care for cancer patients often requires a customized, evidence-based approach to treatment due to the unique characteristics of the disease. Watson for Oncology offers great potential to help enable the best possible patient outcomes and is ideally suited to help advance China’s efforts to improve the quality and efficiency of cancer treatment.”

This initiative is seen as an indication of the rising momentum amongst health care professionals to advance cancer care using cognitive computing platforms.

InvitroCue- A start up that focusses on digital pathology of Liver cancer

A Singapore based biotech company, InvitroCue, offers image analytics services for liver disease applications via a platform that incorporates feature extraction, artificial intelligence and machine learning for more accurate and faster disease diagnosis. The platform, Cuepath, converts images derived from pathology glass slides, CT and MRI scans of liver cancer and fibrosis samples to digital images, and subsequently archive and transmit to various other pathologists.

“If you look at  pathology, there is a lot of variability when different pathologists read the information, so through image analytics we do feature extraction on these digitised images and use machine learning algorithms to analyse them so that you get consistent readouts for any pathological sample and this way there will not be any inter-observer variation at all. This could expedite preclinical research outcome” said Co-founder Dr Steven Fang.

Similar applications of visual analytics are detection of early stage skin cancer, by IBM research group and automated diagnosis by Zebra Medical Vision to predict and even prevent cardiovascular diseases.

There are still plenty of hurdles for machine learning to become a mainstay in disease diagnosis. Creating massive data banks may cause privacy issues in some countries. Also, rare forms with fewer clinical data may still prove hard to detect. But, this could be a minor problem over time, with machine learning being a continually learning system.