Is “IBM Watson Health Imaging” the Future of Healthcare?
A track record of prior competency that is above and beyond the norm is what hiring managers look for when they recruit “top talent”, as recruiters like to say. Usually “top talents” can command a premium in the marketplace because everybody wants to employ them. We can equate these “top talents” to top quality stocks. You often hear dividend investors talk about how top dividend growth stocks are “always too expensive”. That comment usually refers to the yield for the stock being lower than average, which in the case of a quality stock just represents a greater anticipation of future growth. Since many dividend investors value the income from dividend stocks, a low yield means a low income, so it’s considered “expensive”. Inversely, a stock with a higher yield may be paying investors “more now in anticipation of lower future prospects” – which is kind of how a lot of people continue to feel about IBM (IBM) and their 4.3% yield.
Avoiding the “Strategic Imperatives” Conversation
When thinking about IBM’s “future expectations of growth”, there’s always the old “strategic imperatives” chestnut you can crack open, except what you’ll keep finding inside seems old and stale – much like the company’s fearless leader. That’s what one side argues anyways, while the other side claims that IBM is slowly realizing success, with strategic imperatives representing 47% of IBM’s revenues over the past 12 months. We’re not here to disagree with either side, but instead to take a closer look at IBM Watson Health Imaging, one of the five areas of focus for IBM Watson Health.
In August of 2017, we asked the question “Is IBM Ready to Dominate Radiology With AI?“, and in that article we talked about the importance of “big data” in determining who will build the best medical imaging algorithms. Sure, there may be over 40 startups working on using artificial intelligence to do medical imaging, but they are all constrained by the availability of quality data. Then, there’s IBM.
IBM Watson Health Imaging
IBM Watson Health has spent billions of dollars on purchasing medical imaging data and the broader medical data to accompany it. For example, after they acquired Merge for $1 billion in 2015, the Watson Health Cloud had 315 billion data points including lab results, Electronic Medical Records (EMRs), genomic tests, clinical studies and other health-related data sources that are supposed to help improve their AI medical imaging algorithms over time. All of this data should provide IBM with some sort of “medical data superset” which should have them way in front of their competition, right?
If you were to use FDA approvals as an indicator of progress in using artificial intelligence to improve diagnostic capabilities in healthcare, look no further than the progress some startups appear to be making in this area. Here are some FDA approvals for AI solutions that detect medical problems that have been announced this year so far:
- (May, 2018) – FDA permits marketing of artificial intelligence algorithm for aiding providers in detecting wrist fractures
Imagen OsteoDetect, a type of computer-aided detection and diagnosis software designed to detect wrist fractures in adult patients.
- (April, 2018) – FDA permits marketing of artificial intelligence-based device to detect certain diabetes-related eye problems
iDx-DR, is a software program that uses an artificial intelligence algorithm to analyze images of the eye taken with a retinal camera called the Topcon NW400.
- (February, 2018) – FDA permits marketing of clinical decision support software for alerting providers of a potential stroke in patients
Viz.AI Contact application, a type of clinical decision support software designed to analyze computed tomography (CT) results that may notify providers of a potential stroke in their patients.
- (February, 2018) – Arterys Receives First FDA Clearance for Broad Oncology Imaging Suite with Deep Learning
Arterys Oncology AI suite empowers clinicians to quickly measure and track lesions and nodules in MRI and CT scans.
The companies behind that list of FDA approvals are not big names, in fact they’re mostly medical imaging startups we’ve come across before. We covered Imagen and Arterys in a previous article on 9 Artificial Intelligence Startups in Medical Imaging. Viz we covered in our article on 12 Startups Diagnosing Medical Images With AI. This is the first time we’re learning about Iowa startup IDx, but they reached out to us right after this article was published to say that they’ve taken in $22 million in funding so far and that their solution is now being used in patient care at University of Iowa Health Care.
Almost as if responding to this news, the IBM Watson Health Executive Overview states “We will seek FDA clearance when appropriate for the Watson Health Imaging platform.” That might be because what IBM is pursuing is a bit bigger than just getting some diagnostic algorithms approved so that they can be used to make decisions. What IBM is trying to sell is a platform that eliminates something called “practice pattern variation”.
The “Practice Pattern Variation” Problem
Practice pattern variation, also referred to as “Unwanted Variation in Healthcare”, refers to the observation that doctors in different physical locations might practice medicine quite differently. One reason is because the data that each doctor considers in order to make a decision varies so much. If both doctors were working with the exact same data set, and considered the exact same variables equally, then they would be more likely to reach the same conclusions. Then there’s the problem of “prior training”, the quality of which will rarely be consistent between doctors. Regardless of the reason for these “variations”, they represent a great deal of opportunity. Take for example breast biopsies. Last week, a peer-reviewed study was published on how breast biopsies are responsible for “annual false-positive expenditures of $2.18 billion“, a paper that was authored by – interestingly enough – a research team at Truven Health Analytics (an IBM subsidiary).
Just like AI algorithms, doctors often make decisions based on prior experiences, of which both length and quality will differ between doctors. Even if the exact same data sets were made available to each doctor, they may still arrive at different conclusions simply because of past experiences. That’s not the case with artificial intelligence algorithms, which can evaluate each case instantly using all available data (structured and unstructured), and then compare it to 1000s of other outcomes in a matter of seconds. That’s what IBM had in mind when they built a “cognitive data review tool” called “IBM Watson Imaging Clinical Review”.
IBM Watson Imaging Clinical Review
In January of this year, HealthCare Business News published a “Q&A with Steve Tolle, IBM Watson Health Imaging’s vice president of global business development and strategy” during which the IBM Watson Imaging Clinical Review platform was described as a “retrospective quality improvement platform that looks for discrepancies in the medical record.” Mr. Tolle goes on to say that while the version they released in April of 2017 had one clinical indication of aorta stenosis (AS), they are “getting ready to release the second version that will cover 306 diseases“. Here’s an example from IBM Watson Imaging Clinical Review which shows how the thing works for an AS patient:
If you look at “step 2” and “step 4”, you might ask yourself “How does the AI know that the EMR ‘problem list‘ needs to be updated?” The answer may lie in the AI algorithm’s ability to read the echocardiogram data taken from “step 2”. Plenty of literature exists showing that machine learning can now be used to read echocardiograms – like this paper – published in 2009 by IBM’s Almaden Research, which talks about how they built an “automatic algorithm to estimate diagnostic measurements from raw echocardiogram video sequences.” Seems like the video images represent “unstructured data” inputs for the algorithms, while the EMR represents the structured data inputs. All this data could help reduce variation, something that appears to be a problem at the moment for AS patients:
The primary goal of the AI algorithms will be to keep the patient’s “problem list” up to date based on the information it receives, and gently nudge everyone in the right direction. The platform has been sold at least once, with Ohio’s TriState shelling out $10 million in April of this year with a go-live date forecasted to be “early 2019”.
The Future of IBM Watson Health Imaging
It’s easy to envision a future world where IBM Watson ends up analyzing all medical images with accompanying EMR data because it can process such a large amount of information so quickly. A human would never be able to consider that much information in their decisions, even if it were readily available. Now, the computer can watch over the human’s shoulder. “Are you sure you want to mark this positive?” it may ask, prompting the case to be put into another workflow where a second opinion is required. It’s those boundary cases where AI will excel, and consequently it will never actually make a decision but simply help others to reach “the best decision to make, given all information currently available”.
That sounds similar to another product IBM plans to release this year targeting radiologists. Called “IBM Watson Imaging Patient Synopsis”, it provides relevant medical record data to radiologists in near real-time according to a product brochure:
IBM Watson Imaging Patient Synopsis intends to be a radiologist-trained cognitive data synopsis tool to efficiently inform confident care decisions. It aims to provide crucial perspective that can be time-consuming to find in the EMR by presenting clinical context from a variety of data sources… in seconds.
Of course, not everyone’s getting excited about this year’s new commercial offerings from IBM Watson Health Imaging.
IBM Watson Health’s Bad Rap
If you keep an eye on financial news, you would hardly be criticized for thinking that IBM is mucking things up. Again. Just recently, IBM confronted rumors of large Watson Health layoffs calling them “a small percentage of our global Watson Health workforce” in an email sent to the Triangle Business Journal. An article by Stat a few days ago titled “IBM Watson Health hampered by internal rivalries and disorganization, former employees say“, followed one published several days earlier titled “IBM’s problems with Watson Health run deeper than recent layoffs, former employees say“, both implying that either things are isht, or someone has an ax to grind. While the answer is probably somewhere in between, there’s no denying IBM Watson has a perception problem at the moment, with some critics calling it a joke in comparison to Google or Amazon.
When considering the potential of Watson Health as a whole, we need to take into account the other four focus areas (Life Sciences, Oncology, Value-based Care, Government). Some of these areas have taken criticism as well, like oncology’s much-publicized unfavorable outcome with MD Andersen. While imaging is just one part of the overall Watson Health business, it’s still an area that we
had have high expectations for, based on the investments they have made which we talked about before. Over a year has passed since IBM Watson Imaging Clinical Review was released, and we’re still waiting to hear about all the growth coming from healthcare, as opposed to security (a key emphasis in IBM’s Q1-2018 earnings remarks.) While those earnings remarks didn’t even mention Watson Health, let’s hope that doesn’t have to be the case for too much longer.
Pure-play disruptive tech stocks are not only hard to find, but investing in them is risky business. That's why we created “The Nanalyze Disruptive Tech Portfolio Report,” which lists 20 disruptive tech stocks we love so much we’ve invested in them ourselves. Find out which tech stocks we love, like, and avoid in this special report, now available for all Nanalyze Premium annual subscribers.