Cancer is still undoubtedly one of the scariest diseases to be diagnosed with. We have, however, seen significant developments in its treatment. In 2014, we saw a significant tipping point where more than half of those diagnosed survived. Having a 50% chance of survival still isn't too great for those facing the disease, however, work is being done to improve this even further.
One of the keys to making sure that more than half of all cancer patients survive has been the effective use of clinical trials to help in the identification and treatment of cancer.
At the centre of this is the use of data and machine learning. This will not be a surprise to anybody who has worked in the area, as, despite the use of data being seen as a relatively new phenomena, the truth is that it has been used in this capacity for over 30 years. It has had significant results too, and has been shown to improve the ability to predict cancer by between 15-20% - a significant improvement over traditional methods.
It can be used in several different ways within the process, and scientists are finding that it is one of the most powerful weapons in the fight against cancer.
One of the most important elements that it brings is the ability to gather significant amounts of information for a diagnosis or prediction of risk. Due to the power of machine learning, it can look at everything from family history and DNA, through to geographical location and lifestyle choices. The ability to mine all of this together to make a coherent and usable prediction makes it incredibly powerful. Not only this, but it helps to create more accurate and useable clinical trials.
It also allows for significantly fewer human errors in the diagnosis and measurement process moving forward. This comes partly from the huge numbers of people who have been diagnosed with cancer. In the US alone, there was an estimated 14 million people living with cancer. each of whom would have had hundreds of test results and images that can be fed into a machine learning system. This creates a huge dataset that can be mined and collated, allowing for incredible accuracy, even in relatively obscure variations.
Companies like Imageanalysis have utilized this data to make their imaging systems more intelligent and considerably more likely to pick up signs of cancer. This kind of technology allows clinical trials to be more accurate for individuals, providing better results, improving the outlook for patients in a clinical setting and allowing for accurate personalized treatments.
It is also being used to identify the drugs that particular patients are more likely to respond to, most recently with Berg, who are hoping to identify the biological makeup of cancer patients who are likely to respond best to their drugs. Berg CEO and co-founder Niven Narain is confident of its success - 'With use of Berg’s Interrogative Biology platform, we will be applying our precision medicine approach where output from this trial will allow us to match patients to this given combination based on their biological profile.'
Essentially the company will be using machine learning to create what they call a 'Molecular Map' to help identify which patients will respond best to their drug treatment. At present, they are planning on enrolling 25 patients with metastatic pancreatic adenocarcinoma, and the trial is going to run until 2019.
We are at an exciting time for cancer research and the development of drugs to further improve the survival rates. The use of machine learning, not only help to discover and measure cancer itself but to help create drugs to fight specific cancers, is going to be key to this.