Better Health, Cancer, featured, Health Trends

How Artificial Intelligence and Machine Learning Will Fight Cancer in 2018

img cover machine learning

If artificial intelligence (AI) is the bringer of hope for cancer survival, then it must be treated with utmost care, paid its due diligence, and served on a platter to those who desperately need it.

In recent years, AI and machine learning have sent waves through the healthcare industry. Yet, one question lingers in the minds of individuals affected by a cancer diagnosis: “Can we expect AI and machine learning to improve cancer diagnosis and treatment in 2018?

First, it helps to consider that the earlier a tumor is detected, the higher the chances of survival. Scientists are constantly looking for ways to catch cancer before it happens. Second, multidisciplinary teams must also show excellence when it comes to offering the most appropriate treatment plans, as this is necessary for survival.

Below, I highlight 5 ways artificial intelligence and machine learning will help people fight cancer in 2018. Based on recent findings, AI and machine learning will:

1. Cause a Bigger Drop in the Human Error Rate in Cancer Diagnosis. Pathologists spend a great deal of time to try to give the right diagnosis to patients. Still, they make errors. Fortunately, in 2016, a team from Harvard Medical School’s Beth Israel Deaconess Medical Center (BIDMC) showed that when pathologists partner with GPU-powered deep learning analysis, the human error rate in diagnosis drops by 85 percent. As AI continues to rapidly transform the way we do medicine, there will be a bigger drop in the human error rate during diagnosis.


2. Identify More Individualized, Evidence-Based Cancer Treatment Plans. This is already happening. Physicians in Memorial Sloan Kettering have already joined forces with IBM’s Watson Oncology to find a common ground for providing personalized, evidence-based treatment options for cancer patients. This is great news as variations in genetic composition cause individuals to react to treatment differently. Personalized medicine and patient-centric healthcare have existed for a while now. We will see more healthcare teams working together with machine learning experts to use algorithms for providing personalized care for cancer patients.

3. Find new drugs and compounds that can fight cancer cells. In 2017, researchers affiliated with Insilico Medicine were the first to propose the idea that a generative Adversarial Autoencoder (AAE) model can identify new “molecular fingerprints” of compounds that can combat cancer. Their study was insightful and relevant as they used AAE output to screen 72 million compounds with anti-cancer characteristics. The AAE model predictions matched compounds that are already proven to fight cancer, as well as new compounds that have great anti-cancer potential. Overall, deep learning overcomes several limitations of other predictive methods like laboratory experiments. Similar studies are needed to improve the accuracy of predicting anti-cancer drugs.

4. Identify more breast cancer cases at risk of misdiagnosis. To be eligible for appropriate breast cancer treatment, physicians have to first assess the level of expression of the human epidermal growth factor receptor 2 (HER2). Misdiagnosis occurs in 4 percent of negative cases and 18 percent of positive cases. A study published by Vandenberghe provides evidence that deep learning identifies cases at risk of misdiagnosis.

5. Predict more accurate patient outcomes. We are all different. Every cancer case is unique and outcomes are different. Scientists have been able to combine Association Rule Mining with deep learning neural networks (DLNNs) to better predict the outcome of anti-cancer drugs. Their methods outperformed the current predictive gold standards, i.e. Random Forests (RF) and Bayesian Multitask Multiple Kernel Learning (BMMKL) classification. Additionally, deep learning has the power to validate other methods. In 2017, another study used convolutional neural networks and Long Short-Term Memory networks to validate the predictive power of the colorectal tumor tissue microarray (TMAs). The power of machine learning here lies in better predicting more accurate patient outcomes and more accurate responses to treatment.

AI and machine learning are rapidly transforming medical diagnosis and treatment. It is the hope of many that the result of this inside-out transformation will be the death of cancer.

However, Rome was not built in a day. Scientists, researchers, physicians, and individuals need to take it one day at a time. I believe in hope. I believe in life. I believe we can expect AI and machine learning to improve cancer diagnosis and treatment not only in 2018 but in years to come.