Issue #11: Tech is flooding healthcare, but nothing comes close to human professionals
But robots, chatbots, and AI have the potential to disrupt healthcare
Hello there!
Few sectors are as impactful as the medical sector. Doctors, nurses, and hospital administrators make decisions that can impact patients’ lifespans. Research suggests that understaffing, poor access to healthcare, or simply poor quality due to stress is linked to higher death rates.
I have seen several times that overworked doctors are prone to mistakes and wrong diagnoses in my family. Some of my family members ended up in hospital after weeks of waiting for appointments and quick checkups that didn’t find the root cause of their medical issues.
Technology could relax the tense work lives of healthcare professionals by performing time-intense routine tasks and subsequently improve overall healthcare. But where does medical artificial intelligence stand? How will robots change surgeries? I have dedicated this issue to those questions.
There are significant advances in the medical sector, but will patients embrace robot doctors and chatbots, or will they push for more personal care?
Leave a comment to share your thoughts!
Best,
Alice
Headlines you shouldn’t miss
EURONEWS COVID cure? This robot scientist may help Swedish scientists find better treatments for the virus: A robot scientist at the Chalmers University of Technology in Gothenburg is trying the right chemical to fight Covid. The AI-driven system called Eve allows researchers to focus on more demanding tasks and is set up to search for a new compound as a Covid cure.
MEDCITY NEWS AI-driven Deep Genomics gets $180M to turn biology into informational medicines: Deep Genomics CEO Brendan Frey calls the company’s technology a ‘genetic patch.’ The startup is attempting to develop novel drugs based on genetic data.
THE HINDU Robotic cardiac surgery could be the norm in future, says expert: Robotics could drastically improve heart surgeries, according to healthcare practitioners. They would reduce the costs and lower the risks for patients with heart issues. Instead of risky open-heart surgery, robotics can treat cardiac diseases in a less invasive manner.
CNBC Elon Musk’s brain computer start-up raises $205 million from Google Ventures and others: The total amount of investments for Elon Musk’s company Neuralink has $363 million. Musk believes that brain chips will be able to transfer thoughts and memories to digital storage.
DNA INDIA IVF adopts Artificial Intelligence to select the best embryo: Embryologists could receive assistance from an AI-driven system to help them select the embryo with the highest likelihood to become a healthy baby. Experts believe algorithms could quickly scan if an embryo is prone to health conditions. This opens the debate for the ethical use of AI and the future of ‘design babies.’
Covid-19 was the first big test for predictive AI — and the technology failed
When Covid-19 hit Europe in early 2020, doctors didn’t know how to manage the novel virus. They didn’t know how to treat the symptoms or understand the severity of patients’ state. In theory, AI could’ve helped. With countless lung scans available, algorithms could train to identify severe Covid cases and help doctors make the right treatment decisions. In a sobering piece by William Douglas Heaven, the limitations of AI become painfully obvious.
Various studies show that out of the hundreds of AI tools for Covid diagnostics, not a single one was deemed fit for clinical use. The idea of the tool developers was simple: Using machine learning, algorithms should learn certain patterns of medical scans that indicate Covid.
Instead, the algorithms learned how to spot children among adults or individuals lying down instead of standing. In one case, an algorithm learned to spot a font used by a certain hospital for labeling the scans.
Later analyses indicated that the data sets delivered for training the algorithms were flawed, too. Doctors who were not familiar with Covid labeled some scans — their judgment of Covid was enough for labeling scans and not a reliable PCR test.
The experiences of the last year and a half prove that developing a reliable AI that can help healthcare professionals requires more advancements and better data. At least soon, healthcare professionals will have to rely on their own judgment in many cases.
Surgeons might see inside of you with VR goggles
Vicarious Surgical is one of the hottest players in the field of healthcare robotics. Even Microsoft founder Bill Gates has invested in the startup that has developed tiny robots with miniature arms and a camera that allows surgeons to look closely at a patient’s abdomen. Virtual reality (VR) goggles allow surgeons to see where the source of discomfort and pain lies.
The goal is to make abdominal surgeries minimally invasive. The small robot won’t require much space but allows for maximum impact once moving through a patient’s belly. Vicarious Surgical hopes to enter the market in 2023.
The startup wants to specialize in treating hernias for now, but with more experience, it could easily expand its scope.
For surgeons, work is likely to change drastically in the future. It’s not only Vicarious Surgical that tries to develop new robots with more precision than a human hand will ever have.
In the future, a large number of routine procedures could be conducted by robots — if the patients allow it.
Medical chatbots can help identify symptoms, but they aren’t mature enough to make reliable predictions
Last year, the Covid-19 pandemic boosted the use of medical chatbots. A few years earlier, they were serving as hospital receptionists that directed patients to the right departments. But with more insecurity about what Covid symptoms actually look like, chatbots were flooded requests. People wanted to know if they should be concerned about their runny noses or if their symptoms indicated a common cold.
Symptom checkers have functioned like classic decision trees. Answer one question with “yes” and you will continue on another branch with a follow-up question. But with modern AI, the models become more sophisticated. This, however, doesn’t mean that AI algorithms are going to replace doctors any time soon.
First of all, the obvious problem is data: With common conditions like appendicitis, algorithms receive many data points to learn. But the more dangerous cases tend to be rare — therefore, there are fewer data points. In these cases, an algorithm will advise seeing a doctor.
Second, diagnosing a patient is incredibly complex. Imagine you enter a room, and a patient is lying on a bed. As a human, you intuitively perceive various indicators that give you information about a person’s state: You see if the skin has a rash or if somebody is pale. You see if they can open their eyes or are too weak to do so. You see if their lips are dry. And that’s before you’ve spoken to the patient and asked for specific information. A trained healthcare professional sees even more indicators and all the cues combined help to make a diagnosis. This is incredibly hard to emulate.
Soon, AI could assist in checking basic symptoms, but with Covid-19 becoming more manageable, patients are likely to seek human advice.
Opinion: AI is unable to tell why it takes a certain decision — and it’s problematic
Researcher Jason H. Moore shares his thoughts on AI in healthcare in ‘Scientific American’. While AI has been trained to scan and identify medical conditions like eye disease, there is one practical problem: the algorithms can’t explain how they came to their conclusion.
Global institutions World Health Organization demand that AI algorithms shall be transparent, unbiased, and explainable. However, when corporations own the algorithms, it might be difficult to have transparent access to the way they operate.
Additionally, algorithms have become too complex for humans to understand fully. One of the first healthcare algorithms was developed in the early 1970s, and it indicated whether a patient should take antibiotics or not. Developers could track its decision tree and the probabilities for its recommendation. With modern AI, tracking decisions becomes nearly impossible.
Doctors could find themselves in a situation where they receive recommendations they can’t follow. This might not only jeopardize the health of patients but create legal issues around accountability.
The optimistic outlook: AI could soon deliver personalized treatment
The good thing: The authors don’t speculate about what could be, but they name companies working on modern healthcare. For instance, they refer to the Finnish company Oura as an example of personalized healthcare.
They name Insilio Medicine as one of the companies that could accelerate drug development and see PathAI as a key player in more reliable cancer diagnostics.
Diamandis and Hsu believe that the future of healthcare will be more precise, personalized, and reliable, thanks to the convergence of AI, robotics, and new technological advancements.
Tweet of the week: The EU Commission’s plan for the future
The European Commission dedicates a lot of money and resources to its NextGen package to foster a prosperous future. One of the aspects in the package is AI in healthcare: