AI in Medicine – A Historical Perspective
Artificial Intelligence (AI) is going through a renaissance, fueled by powerful deep learning algorithms and generous amounts of investment over the last several years. Sure enough, various AI technologies are listed at or around the “peak of inflated expectations” on the Gartner hype cycle for 2017. VC investments in AI companies has reached historical peaks, with over US$ 5 Billion invested in venture-backed companies in 2016.
At RowAnalytics, we’re looking forwards towards the future potential of AI in precision medicine, but a historical perspective always affords a lesson learned. AI veterans would recall that this is not the first hype cycle for AI (see Figure 1). Early AI research efforts ramped up through the 1970s and into the 1980s. The initial hype cycle peaked with ambitious projects like the Fifth Generation Computer project in Japan and in the Cyc project in the USA in the 1980s, followed by a long and arduous “AI winter” in the late 1990s when research funding had all but ceased.
Successful AI Solutions Augment Human Skills, not Replace Them
Throughout the history of AI, medicine has been a popular topic among AI researchers. AI in Medicine efforts started in the 1970s with efforts to automate diagnosis. These efforts produced some interesting results but without much acceptance in the medical community. At the time, one of the initial observations was that doctors could not trust AI systems since these systems could not explain how they reached their decisions. In the 1980’s, the community shifted its attention to AI systems that explained their findings. Regrettably, the ability to explain results did not make an impact for acceptance and use of these systems.
In retrospect, one might surmise that people did not have a burning need for AI assistance with cognitive tasks such as diagnosis and planning (several successful projects and limited deployments nonwithstanding). Back then, researchers realized that diagnosis is not only the primary work product of doctors, but something they did rather well and didn’t need help with. That remains true today. Besides, there’s the whole issue of liability for a wrong diagnosis or treatment. With a doctor in charge, the chain of accountability is well defined. With AI, not so much. Therefore, most successful AI efforts in medicine to date have been with technologies that augment human skills. Conversely, solutions that aim to replace humans with AI have not fared well in medicine, in general.
Cognition or Classification?
Three decades later, the focus of AI has largely moved from higher cognitive skills to classification, a lower-tier cognitive task. The marriage of autonomous driving with classifiers based on deep learning quickly moved AI from R&D labs into daily life. Still, it is important to understand the current limitations of classification and pattern recognition technology. Image recognition still needs to go a long way, for instance, to identify a video from a random high school play as a musical rendition of Cinderella. Humans are remarkably good with such interpretations whereas machines are not that robust (yet).
What about the ability of an AI system to explain its reasoning and decisions? Perhaps people do NOT need explanations from an AI system – they just need to be convinced that the results are accurate, consistent, complete, and repeatable. It is important for the decisions to be robust and the failures graceful and gradual. For that, the domain may need to be sufficiently constrained so that an AI system is never required to classify and identify something that it has not seen before.
Today, some AI image recognition systems exhibit remarkable accuracy in detecting and diagnosing melanoma. They are more accurate than non-expert physicians already, and definitely more consistent than a panel of human experts. Accurate and consistent decisions on suspicious-looking moles would be sufficient for most everyone, but they may still need to be convinced that the AI system is able to differentiate skin lesions from coffee stains on a carpet! Humans can; AI classifiers may or may not.
Has the Time Come for the Robot Surgeon?
Another interesting potential application area for AI in Medicine is surgical robotics. Certainly, there are those who predict that a majority of surgeries would be conducted by robots within a couple of decades. I am not as optimistic. The state-of-the-art today is robot-assisted surgery. It will be a long time before AI is allowed to operate on humans without human participation or at least oversight. Most of the trivial aspects of surgery (or, for that matter, flying a 777 across the Pacific with three hundred people on board) may be performed by robots today. What makes human expertise indispensable is management of emergencies. In surgery, I have seen situations where certain internal structures are not where you expect to find them due to anatomic variations or impact of disease (e.g., a tumor displacing everything around it). Every so often a small blood vessel gets nicked and there’s blood everywhere, obstructing visual cues and requiring fine finger manipulation to locate the bleeder. When working near the intestinal tract, a sense of smell is useful to have!
Having said all that, there might be limited instances where surgery may be performed safely by AI on humans today. One example that comes to mind is Mohs surgery for certain skin cancers. This is a kind of surgery that requires precise excision of the cancer followed by a rapid microscopic analysis of the sample while the patient awaits on the operating table, and enlarging the excision progressively until no traces of tumor are found on the excised tissue sample. Another example is LASIK eye surgery, which, for all intents and purposes, is already done by a robotic device under the close supervision of an eye surgeon. Today, the eye surgeon is still responsible for all measurements, adjustments, and calibration of the device, but it’s safe to assume that more of those functions will be handed to AI in the near future. The use of teleoperated robots is also expanding and giving hope to patients who cannot be served otherwise (e.g., battlefield or remote locations). As far as robots taking over from human surgeons entirely, that requires robots reaching a level of consistency and reliability that’s on par with the average surgeon – and similar skills to handle and manage emergencies and unknowns. We’re not there yet, and that might take another fifty years.
The Future of AI is in Precision Medicine
In our own line of work at RowAnalytics, we see valuable applications of AI technology in the service of precision medicine. Our AI algorithms are able to discover previously-unknown genetic pathways that are implicated in certain complex diseases. We have semantic search technologies that perform much better than simple keyword-based searches on complex queries. Our AI methods also fuel personalized diet, lifestyle, and health advice for patients with complex healthcare issues. As someone who has been involved AI in medicine for almost three decades, I am delighted to see AI in medicine technologies finally mature and move out of research laboratories into daily life.