When it comes to effective medical treatment in colorectal cancer (CRC), early detection is key. Colorectal cancer (CRC) is the third most common form of cancer found in men and women in the United States, and is estimated to be responsible for around 52,000 deaths each year. Despite those grim numbers, colorectal cancer is also highly treatable -- particularly when it is discovered early. And the earlier the better.
Researchers in Israel have developed a machine learning model that’s able to assist with early identification of colorectal cancer risk; identifying possible patients six to twelve months prior to current diagnosis. Such colorectal cancer screening could potentially reduce the mortality rates for this disease by between 15 percent and 30 percent. That equates to thousands of people each year whose lives could be saved by such innovation.
Currently, screening for colorectal cancer may be carried out through screening colonoscopy or checking occult blood in stool. Both methods require the patient to be very active so the compliance is low.
The medical startup Medial EarlySign, working with Maccabi Healthcare Services, has found a way to screen for colon cancer passively only by looking for biomarkers which can suggest the possibility of colorectal cancer or even a pre-colon cancer state. This can be done in a non-invasive, passive manner that utilizes smart algorithms to carry out scans of patient health record data including existing lab tests, and then -- where appropriate -- alert physicians when their patients might potentially be at risk. Unlike symptom-based models, which tend to be usable only at later stages of the illness, this tool could be used early in the process, prior to observable symptoms emerging.
The tool is likely to get even better at prediction, as more and more data is fed into it. In the same way that Netflix can predict the shows that will most appeal to you, based on the responses of other users with similar tastes, so too will tools such as this get smarter as more data is added.
Maccabi is now taking information from complete blood count which is the most prevalent blood test to see if they contain biomarkers which could serve to indicate early warning signs of colorectal cancer. The innovation is that passive screening on your existing EMR data can diagnose you early. The ability to create real, valid risk prevention models is therefore a huge medical revolution in the making that can be applied to many different conditions -- saving time, money, and even lives in the process.
This kind of approach to precision medicine is built around what are termed patient similarity networks. A patient similarity network is an analytical, data-driven means by which to “cluster” or classify different patient profiles together based on similarities in various areas, such as genomic profile. Using electronic medical records (EMRs) covering everything from blood samples to age and gender, it’s possible to tap into the power of crowdsourcing to seek out fresh insights that can help individuals.
Clustering is important because it connects the “nodes” of individual patients into a large data set that can be tapped for insights -- the same way each individual Google user clicking links makes the overall system smarter. It’s a fascinating observation that it’s this connecting of large groups of patients, complete with a sprinkling of machine learning magic, that is making truly personalized medicine possible. As electronic medical records become more widely available to patients around the globe, tools like the one above will only become more commonplace.
Right now, we’re still at the start of this particular journey. But a few pioneering companies are already showing the kinds of benefits we can expect -- and, while many are related to physical patient outcomes, there are other advantages too. At Alike, we use a patient similarity network to help users find their medical “Alikes,” meaning others who share their profiles. When two individuals match highly enough on a similarity score, calculated using Alike’s proprietary algorithms, they can then be connected. Patients can then share their insights, such as information about what they’ve learned on their medical journey that may provide help in better managing and coping with different conditions. By doing this, patients are not only supported medically, but also socially -- with an empathic partner who, quite literally, has been through the same experience.
The world is still at the beginning of this personalized medicine revolution. But it’s already shaping up to be transformative. If it can save lives, it could prove to be even more than that.
.png)
.png)
