Technological advancements are things that the medical field relies on heavily. We all know the usual medical technologies such as x-ray machines and heart rate monitors, but what most people don’t know is that technology like artificial intelligence (AI) is changing the game for doctors and patients; specifically cancer patients.
According to a McKesson Corporation Survey, the World Health Organization states that nearly 10 million people will die from cancer this year and more than 18 million cases will surface around the world. Although most doctors try their best to give the most accurate and up-to-date diagnosis and treatment options, there are limits to the diagnostic abilities of doctors, and with further data analysis, a more in-depth and accurate option may be on the horizon. This is where AI comes in.
Artificial Intelligence is made possible through the use of algorithms. Engineers who make AI technology actionable do so through adding data into an open-source algorithm, which allows the system to learn the information and make inferences from that data. This is exactly what scientists from NYU did with Inception-v3.
In order to make the AI useful, NYU scientists took an off-the-shelf Google deep learning algorithm and trained it to differentiate between different types of lung cancer with 97 percent accuracy. The type of Google AI system that they used is the same used for reverse image searching and can identify objects in an image. This technology has been used in the past to diagnose diabetic blindness and heart conditions, but this was one of the first times it has been used to identify genetic mutations.
The team first trained Inception-v3 to identify 1,000 different types of images from the Cancer Genome Atlas to distinguish between cancerous tissue and healthy tissue. Once it learned the difference, it was given images of the two most common types of lung cancer cells, being adenocarcinoma and squamous cell carcinoma. It learned the difference between how these cells look and was able to identify between healthy tissue, cancer tissue, and two different types of cancer cells.
While Inception-v3 was being built in the U.S., a team led by the Institute of Cancer Research London and the University of Edinburgh in Scotland were busy developing a new AI technique called a Revolver or the Repeated Evolution of Cancer. With this new technology, the team taught could pick out the patterns in DNA to find the genetic mutations of cancer cells. In pinpointing the exact mutation that caused a cell to be cancerous, doctors are now able to compare one patient’s cancer to another who’s had the same or a similar mutation to forecast for future treatment.
One of the biggest challenges in cancer treatment is the genetic mutations of cells which cause the cancer to be drug-resistant, however AI technology oncologists now have an unlimited dataset that can help them get ahead of the cancer by providing them recommendations for personalized treatment. For example, researchers found that breast tumors that had genetic errors that coded for the tumor-suppressing protein p53 along with mutations in chromosome 8 have a lower prognosis and survival rate than those with similar genetic mutations without chromosome 8 mutations. Information like this is life changing to so many breast cancer patients, as well as other types of cancer patients whose cancer stemmed from genetic mutations.
Many of the cancers that are being helped by AI have to do with genetic mutation, but what about the rarer cancers that involve no genetic mutation at all? Don’t worry. AI is targeting them as well. Canon Medical Research Europe was recently awarded $180,000 to develop an AI prototype that can recognise and asses asbestos-related cancer tumors that go along with mesothelioma, a cancer of the lungs. Under the Cancer Innovation Challenge in Scotland, a competition that aims to make Scotland a world leader in cancer care, the team was given funding to develop this technology.
Ken Sutherland, president of Canon Medical Europe, told the Scottish Business Insider, “Malignant Pleural Mesothelioma is a terrible condition for those that are unfortunate enough to suffer from it, and we believe that an automated assessment method using AI would be a major advance in fighting this disease and, potentially, other forms of lung cancer.”
Concerns with Ethics
A McKesson Corporation survey of more than 2,000 participants reports that 60% of respondents say they are open to genetic testing when it comes to assessing their risk in developing cancer, however, before we bring AI to real world everyday scenarios, the whole idea needs some ethical evaluation.
According to Stanford Medicine, one of the major concerns is bias reflected in recommendations. Depending on who designed the algorithm, there is a potential for skewed results when testing. Corruption may occur from companies or healthcare systems who are benefitting from certain types of treatment recommendations. Furthermore, physicians will need to understand how the system works, which is possible, but also very complicated due to the vast amount of data the algorithm requires to work properly. However, those who work with algorithms should be warned to understand them on a base level and not to become overly dependent on them.
Data collection is also an ethical issue that many physicians are concerned about. Collected knowledge about patients is something that would have to be done at mass scale and has the potential to be harbored without patient knowledge. Not to mention, this machine learning would introduce a third party into the doctor-patient confidentiality agreement, which challenges the responsibility and expectation of patient privacy.
Even though there are hurdles to overcome before this technology is used on an everyday basis, artificial intelligence has made major advances in cancer treatment. With the nature of cancer symptoms being so vague and so similar to less crucial illnesses, doctors are feeling the pressure more than ever to give more accurate and timely diagnoses. With technology like AI, this will help doctors provide patients with personal treatment and on their way to beating cancers, rare or common alike.
Guest Post Written by Rachel Lynch