Every industry has buzzwords that fan out across culture and become synonymous with a job well done. The ‘80s business culture brought us “synergy,” which is now used to describe any systems that work together well, albeit sometimes mockingly. During the 1990s, the dot-com boom came and brought with it “enterprise solutions,” and now in 2018, it seems as though everyone is talking about artificial intelligence.
What Is Artificial Intelligence?
The term artificial intelligence (AI) came about in 1956, and since then, AI has progressed greatly. The first advances in AI focused on the construction of neural networks. They were modeled after the human brain's ability to take inputs and produce outputs from the given data. Around the 1980s, these artificial neural networks progressed to a point where “machine learning” became popular.
Machine learning refers to a machine's ability to skim through data and find patterns, thus learning from the data and then applying it to problems to make informed decisions. This kind of AI can be found in malware detectors, services like Spotify that suggest music based on past preferences, and financial software that suggests favorable stock trades.
After machine learning came the trend of deep learning, which is a more sophisticated subset of machine learning that requires no human intervention for the machine to progress. A machine that uses more basic machine learning makes predictions and will still need human guidance if those suggestions are wrong to then make better ones. Machines with more advanced deep learning capabilities can deduce if they've made good predictions on their own and continue the process of learning from these deductions.
AI machines today utilize a mixture of machine learning and deep learning, and these machines can be applied to a vast array of disciplines.
AI in Cancer Detection & Diagnosis
In the health services industry, AI has a wide range of uses and applications, from helping with clerical work to turning the tide in the race against cancer. Any oncologist knows that early detection of a cancer is necessary for the successful treatment of malignant tumors.
Tumors inside of a patient's body are most commonly detected through medical imaging techniques, such as radiology. Radiologists today receive more data than they can humanly work through in one shift. A study published in Academic Radiology showed that an average radiologist must interpret one image every 3-4 seconds to keep pace with their daily workload (2015;22(9):1191-1198).
AI components in imaging machines would reduce this workload and drive greater efficiency in the radiology field. Machines also have access to a greater wealth of data than their human counterparts do, which can mean that an AI machine can detect cancer with more accuracy than a human.
A deep learning algorithm developed by engineers at University of Central Florida's Center for Research in Computer Vision, was very successful in accurate detection of lung cancer. The engineers fed 1,000 CT scans to the AI to teach it how to analyze lung tissue for abnormalities.
This study found that AI machines could identify lung cancer from a scan 30 percent more accurately than humans. Ulas Bagci, PhD, Assistant Professor and group lead of the researchers, is excited about this level of success.
“I believe this will have a very big impact,” Bagci told UCF Today. “Lung cancer is the No. 1 cancer killer in the United States and if detected in late stages, the survival rate is only 17 percent. By finding ways to help identify earlier, I think we can help increase survival rates.”
For some cancers, survival rates are incredibly bleak, so AI could be the catch-all solution many patients hope for. Mesothelioma cancer, which has a 5-year survival rate of 12 percent and a 10-year of less than 5 percent, is particularly deadly.
A Scottish initiative to create cancer treatment technology breakthroughs called the Cancer Innovation Challenge awarded £140,000 to Edinburgh-based firm Canon Medical Research Europe to work to improve malignant pleural mesothelioma assessment with AI.
Besides lung cancer and mesothelioma, researchers around the world are developing ways to detect and diagnose breast cancer, colorectal cancer, and neck tumors, to name a few. Given the expanding rates of research in this field, it seems the potential for AI in oncology is vast.
AI & Cancer Treatment
Research in the potential for AI to treat cancer is less progressed than in detection and diagnosis. Several forays into treatment planning have been made by big-name computing researchers like the IBM Watson team and Google's DeepMind division.
An AI tool designed by a University of Toronto team has shown promise in reducing the time to tailor radiation treatment plans to individual patients. This particular AI used historical radiation data to recommend treatment strategies with success comparable to human radiologists.
For human specialists, it can take days to develop these plans, costing time for both the radiologist and the patient. In 20 minutes, the Toronto team's AI was able to replicate complex treatment plans arrived at by top radiologists.
Google's DeepMind is also designing AIs with comparable purposes. Currently, DeepMind's venture into cancer treatment is in its infancy, though it can recommend treatment for certain eye diseases with 94 percent accuracy. A particularly useful feature of this system is the ability to tell doctors why these suggestions were made.
Instead of simply solving for x and relaying to doctors the most effective treatment, this system gives explanations, reasoning to back the various options, and even confidence scores for suggested treatments. This can help not only with doctor learnings but with fine-tuning the system itself.
In the vein of treatment planning, DeepMind has noted that “early results from our partnership with the Radiotherapy Department at University College London Hospitals NHS Foundation Trust suggest that we are well on our way to developing an artificial intelligence system that can analyze and segment medical scans of head and neck cancer to a similar standard as expert clinicians. This segmentation process is an essential but time-consuming step when planning radiotherapy treatment.”
However, one of the biggest names in AI and supercomputing, IBM's Watson, is still struggling with oncology applications. Touted by IBM as the most promising option for a cure for cancer, Watson still stumbles over basic tasks. STAT News published an article last year reviewing Watson's abilities and commented, “contrary to IBM'S depiction of Watson as a digital prodigy, the supercomputer's abilities are limited.”
Documents reviewed by STAT News from a partnership with doctors at Memorial Sloan Kettering Cancer Center revealed that Watson still has difficulty distinguishing between types of cancer.
In addition, this computer gave incorrect suggestions in the treatment cases it was given. When given a hypothetical case of a cancer patient with severe bleeding, Watson suggested a drug that would exacerbate the bleeding.
Perhaps Watson's downfall came from IBM's overhyping of the product, or maybe training on synthetic cases isn't doing the computer any favors. In any case, using AI to tailor treatments for cancer patients is not a market-ready solution quite yet.
Undeniably, AI has immense potential in the medical field. For cutting down on workloads and streamlining a physician's day, it may be the best man for the job. However, it's certainly not yet at the stage where it can be considered a miracle.
Even successful AIs, such as the University of Toronto's radiology tailoring system, aren't meant to be used without supervision. Author of the Toronto team's study, Aaron Babier, PhD, stressed that their AI wasn't designed to replace radiologists, but rather to lessen the workload of a radiologist and make informed suggestions to be reviewed by a technician or physician.
AI may one day be an exciting option for doctors everywhere, but currently it's just that: in the future.
This article was authored by staff the Mesothelioma Cancer Alliance. This organization serves as a source of information, support, and community for those affected by mesothelioma cancer and asbestos-related diseases.