Artificial intelligence programs are capable of improving healthcare in a variety of different ways. For instance, AI applications can use computer vision to help doctors diagnose conditions from X-rays and FMRIs. Machine learning algorithms can also be used to help reduce false-positive rates by extracting subtle patterns from data that humans may not be able to find in medical data. However, with the possibilities comes new challenges, and recently a new article was published in Science that examined possible risks and regulatory strategies for medical machine learning techniques in an effort to minimize any possible negative side effects of employing AI in a medical context.
Expanding Applications For AI In Healthcare
AI is seeing its applications in the medical field expand rapidly. Recent developments in the field of healthcare, driven by AI, include the creation of a new pharmaceutical company that aims to use AI to create new drugs, the creation of AI-drive remote health sensors, and computer vision apps that analyze CT scans and X-rays.
To be more precise, Genesis Therapeutics is a startup that is aiming to use AI to speed up the process of drug-discovery, hoping to create drugs that can reduce the severity of debilitating diseases. Genesis Therapeutics is just one of almost 170 different firms using AI to research new drug formulations. Meanwhile, in terms of health monitoring devices, iRhythm and French AI startup Cardiologs are making use of AI algorithms to analyze EEG data and monitor the health of those who have heart conditions are at risk of complications. The software designed by the companies can detect heart murmurs, a condition caused by turbulent blood flow.
Finally, a recent study investigating how computer vision can be applied to medical images found that computer vision systems perform at least as well or better than expert radiologists when examining CT scans to find small hemorrhages. The algorithms used in the study were able to render predictions after examining CT scans for just one second. The computer vision systems were also able to localize the hemorrhage within the brain.
So while the potential benefits of using AI in healthcare are clear, what is less clear is what new challenges and risks will arise as a side-effect of employing AI within the healthcare field.
Regulating An Expanding Field
As TechXplore reported, in order to assess potential drawbacks of using AI in healthcare, a group of researches recently published a paper in Science, aiming to derive answers to anticipate potential problems with AI and explore potential solutions to these problems. Problems that may arise from using AI in the healthcare field include the inappropriate recommendation of treatments resulting in injury, privacy concerns, and algorithmic bias/inequality.
The FDA has only approved medical AI that uses “locked algorithms”, algorithms that reliably produce the same result every time they are run. However, much of AI’s potential lies in its ability to learn and respond to new types of inputs. In order to enable “adaptive algorithms” to see more use and get approval from the FDA, the authors of the paper took an in-depth look at how the risks related to updating algorithms can be mitigated.
The authors advocate that machine learning engineers and researchers should focus on continuous monitoring of models over the lifetime of their deployment. Among the suggested tools to monitor AI systems was AI itself, which could help give automated reports on how an AI is behaving. It’s also possible that multiple AI devices could monitor each other.
“To manage the risks, regulators should focus particularly on continuous monitoring and risk assessment, and less on planning for future algorithm changes,” said the authors of the paper.
The authors of the paper also recommend that regulators focus on developing new methods of identifying, monitoring, assessing, and managing risks. The paper applies many of the techniques that the FDA has used to regulate other forms of medical tech.
As the paper’s authors explained:
“Our goal is to emphasise the risks that can arise from unanticipated changes in how medical AI/ML systems react or adapt to their environments. Subtle, often unrecognised parametric updates or new types of data can cause large and costly mistakes.”
How Governments Have Used AI to Fight COVID-19
Governments all around the globe are using artificial intelligence (AI) to help fight against the ongoing COVID-19 pandemic. The technology is being used for various different things, including speeding up the development of testing kits and treatments, giving citizens access to real-time data, and tracking the spread of the virus.
Here are some of the different countries.
South Korea’s government, one that is being touted as an example for how to combat the virus, pushed their private sector to start developing testing kits right away, immediately after the reports began to arrive out of China.
One of those companies was Seoul-based molecular biotech company Seegene, which used AI to help quicken the process of developing testing kits. The company was able to submit its solution to the Korea Centers for Disease Control and Prevention (KCDC) just three weeks after the scientists began their work. According to Chun Jong-Yoon, founder and chief executive of the company, the process would have taken at least two to three months without the use of AI.
Since the testing kits are such a crucial part of getting the virus under control, this proves that AI technology can play a huge role in the fight.
Traditionally, the approval process for new medical equipment, including testing kits, takes around 18 months. The KCDC decided to move the process ahead and approved the tests in one week. The government’s own patient samples were able to be used for evaluation.
Telecoms firm KT has also partnered with South Korea government ministries in order to use AI-based healthcare services to track the spread of the virus.
The research project was led by the ICT Ministry and Ministry of Interior and Safety, along with universities and research institutes. KT will be responsible for providing mobile data, who can help create maps. These maps can then provide insight into how the populations are moving and the virus spreading.
Scientists in China used AI in order to speed up scientific processes. Through the use of the technology, they were able to recreate the genome sequence of the virus in a month. Comparing that to the months it took scientists to create the sequence of the SARS virus in 2003, it is a big step up.
In Taiwan, the technology has been used as well. Audrey Tang, Taiwan’s digital minister, relied on AI to develop real-time digital updates. These updates could alert citizens of certain hazardous locations, where infections had been previously detected. They were also able to use it to create a live map of local face mask supplies.
In the United States, the White House Office of Science and Technology Policy called for the development of the COVID-19 Open Research Dataset (CORD-19). This dataset is a collection of thousands of different machine-readable COVID-19 literature. There are over 44,000 scholarly articles that can be used by the research community.
“It’s all-hands on deck as we face the COVID-19 pandemic,” Microsoft’s chief scientific officer, Dr. Eric Horvitz said. “We need to come together as companies, governments, and scientists and work to bring our best technologies to bear across biomedicine, epidemiology, AI, and other sciences. The COVID-19 literature resource and challenge will stimulate efforts that can accelerate the path to solutions on COVID-19.”
While there is still a lot more that can be accomplished with artificial intelligence (AI), these are some of the current examples from around the globe. If governments are convinced by the results, the use of AI during a pandemic could become one the first defense options in the future.
Anton Dolgikh, Head of AI, Healthcare and Life Sciences at DataArt – Interview Series
Anton Dolgikh leads AI and ML-oriented projects in the Healthcare and Life Sciences practice at DataArt and runs educational and training developers focused on solving business problems with ML methods. Prior to working at DataArt, Dolgikh worked in the Department of Complex Systems at the Université Libre de Bruxelles, a leading Belgian private research university.
What was it that originally inspired you to pursue AI and life sciences as a career?
A passion for searching the links between phenomena and facts. I always like to read. I love books. At university, I discovered a new source of information – articles. At some point, it appeared that to get a complete picture, to crystallize the beautiful truth from a mass of information is almost impossible. And here comes AI. Statistics, machine learning of course, and natural science with AI at the top all act to build the bridge between the human brain’s thirst for knowledge and a world where all the laws are known and there are no black boxes.
You currently educate and train developers who are focused on solving business problems with ML methods. Is there a specific field of machine learning that you focus on more, for example deep learning?
Yes, deep learning is a very popular and, let’s be honest, powerful instrument; we cannot neglect it. I personally prefer the Bayesian interpretation of classical algorithms, or even a combination of neural networks and a Bayesian approach — for example, a Bayesian Variational Autoencoder. But I believe that the most important thing to teach new ML guys is not to blindly use ML machinery like a magic black box, but rather perceive the basic principles behind each and every method. A must-have skill is the ability to explain the predictions obtained for a business audience.
In March 2019, you wrote an article called ‘Are we Ready for Machine Radiologists, and their Mistakes?‘. In the article you outlined the pros and cons of accepting results from machine radiologists versus human radiologists. If you had to choose between a human and a machine giving you results, which would you choose and why?
I prefer a human radiologist. Not because I have some special knowledge that AI tis heavily prone to errors and decisions are intrinsically erroneous. No, it’s more a question of empathy and the psychological nature. I want to support human doctors during this difficult period. Moreover, I believe that in the nearest future, we will only see AI augment human ability.
You recently wrote a white paper called ‘The Impact of Artificial Intelligence on Lifespan.’ In this paper, you stated that AI should be viewed as a tool in the search for longer life. What are some of the more promising methodologies that AI could apply to the quest to extend human lifespan?
Today the new tool of AI is beginning to operate in scientific laboratories on par with classical instruments and approaches. This fact itself is promising. AI is here to help, not replace, us in the struggle to cope with the huge quantities of data flooding not only laboratories but even our personal lives.
Also discussed in the same white paper is a claim by Biogerontology Research Foundation AI Director and CEO of Insilico Medicine, Dr Alex Zhavoronkov, that increasing lifespan to 150 years is not a fantastic goal. Do you believe that a child born in 2020 will be able to live to 120 or even 150 years?
I want to believe. Being a scientist by education and belief, I have to base my decisions on facts, on understanding the progress of scientific methods in the area. We’ve made an impressive leap in the fields of genetics, biotechnology and medicine in general, and this strengthens my belief. And don’t forget that a substantial part of the success in increasing lifespan is a healthy environment and a healthy lifestyle, so we have to work on this.
In this same paper you mention the potential for mind uploading (transhumanism). Do you believe that this could eventually be a reality, and how does it make you feel personally?
I’ve thought a lot about it. Frankly, it makes me feel frustrated. I think that we associate personality with what we see in a mirror, and for me, it’s hard to detach my character from my body. Nevertheless, this doesn’t mean it is not possible. And, yes, I believe that sooner or later mind uploading will become feasible. The consequences are much harder to foresee.
You’re currently the Head of AI, Healthcare and Life Sciences at DataArt. What are some of the most interesting projects DataArt is currently working on?
We have a project dedicated to new drug development. It’s inspiring how computational methods have developed to fuel and direct the progress in medicinal chemistry and pharmacology. We also do a lot of work on applying AI to extract information from medical texts such as clinical trial reports, medical articles, and specialized forums. It’s hard work, but it takes us closer to the digitalization of healthcare, and I find this exciting.
As an avid book lover, I also need to ask what books you recommend?
- Judea Pearl “Causality: Models, Reasoning and Inference“. The title is self-explanatory – the book is about causal relationships. If (one day) we want to have a true AI we have to teach it reasoning about the cause and effect;
- If you’re going to delve into the causative and practical methods then the fundamental work of Daphne Koller and Nir Friedman “Probabilistic Graphical Models: Principles and Techniques“ will be the right choice;
- We expect powerful AI to be able to understand us. Thus, we have to teach the language to it. Natural Language Processing tackles the problem of natural languages’ comprehension. I have two titles in mind that helped me a lot:
- Yoav Goldberg, Neural Network Methods in Natural Language Processing (Synthesis Lectures on Human Language Technologies), 2017
- Christopher D. Manning, Prabhakar Raghavan, Hinrich Schütze An Introduction to Information Retrieval, 2009
- Not sure the next book is dedicated to AI, but it demonstrates a non-standard approach to statistics and predictions which will be useful for any AI researcher: Bertrand S. Clarke, Jennifer L. Clarke Predictive statistics: Analysis and Inference beyond models
- And I would end the list with a sci-fi book: Stanislav Lem, The Star Diaries
Is there anything else that you would like to share about DataArt?
DataArt is an excellent example of the recent trend toward digitalization of almost every aspect of life and activity. This trend increases responsibility in software development because today it’s not only about building a site for a shop, for example, in which case a mistake by the developer will have minimal consequences. Today a developer’s mistake could become a national or worldwide catastrophe if it involves a program controlling the functioning of, for example, a nuclear plant. DataArt’s responsible approach to software development in a broad sense gives me confidence in what we develop, and I am very proud to be part of the company and the work that we are doing.
As for another recent project of hours, last year DataArt launched a prototype application called ‘SkinCareAI’, which analyses skin images to detect early signs of melanoma. Featuring the latest advancements in machine learning (ML) technology, SkinCareAI was developed by DataArt ML expert Andrey Sorokin for the International Skin Imaging Collaboration (ISIC) challenge.
To learn more about some of our other projects and case studies, please go to DataArt’s Healthcare and Life Sciences page.
AI Discovers Smell Genes Linked To Cancer Outcomes
A team of researchers from Oxford University has recently used AI to discover a potential link between colon cancer and the expression of specific smell-sensing genes. As Phys.org reports, researchers from Oxford University and the University of Zurich have recently, assisted by an AI model, discovered that the expression of specific smell-sensing genes within the colon cancer cells indicates a higher probability of worse outcomes.
Genes are expressed when the information that is found within our DNA is used to make molecules like proteins. Gene expression often controls how many proteins are made and when they are made, acting on/off switches. Human beings have approximately 400 genes responsible for our sense of smell, but rather crucially for the study, the genes are also expressed in other parts of the body, aside from the nose. If these smell genes are expressed, it means that the instructions for these genes are being read and carried out. By making changes to the cells, scientists can manipulate the level to which genes are expressed and used.
The study that was recently published in Molecular Systems Biology was lead by Dr. Heba Sailem of the Insitute of Biomedical Engineering and Oxford University. Sailem and fellow researchers examined how cells in the body are organized, aiming to study how cancer leads to the loss of tissue structure in the body. In order to develop effective therapies, scientists must understand which genes play a role in tissue alteration. The research team employed computer vision algorithms to detect changes within the organization of cell samples. The AI model was given image data collected by robotic microscopy, which contains millions of images of colon cancer cells.
The research team then experimented by reducing the expression of every gene in the individual colon cancer cells. After perturbations were applied to the genes and their expression decreased, the researchers found that smell-sensing genes appear to be strongly correlated with how cells align and spread. It appeared that reducing the expression of smell genes could potentially control the spread of cells by reducing their ability to move. On the other hand, cell motility could be increased by having higher expression levels of the smell genes in question.
Sailem explained that the smell genes are like a “sixth sense” that the cancer cells can use to find their way outside of the tumor environment, which is toxic, and spread to other regions of the patient’s body. Sailem went on to explain just how important AI was in making this discovery. The AI model used by the researchers was able to greatly increase the speed the research was carried out at. The AI model, after being trained on a large database of gene functions and appearances, is able to automate the task of identifying certain types of cells within images. Sailem explained:
“Using the developed AI system, we can now learn much more from these experiments and accelerate the identification of genes that alter the structure of tissues in cancer.
CRIPSR (Clustered Regularly Interspaced Short Palindromic Repeats), the gene-editing technology, is the primary way that gene expression levels for the approximately 20,000 genes in the cell are reduced to study how gene expression impacts cancer cells. When combined with advances in gene-editing technology, the research done by Sailem and colleagues could enable new methods of identifying the roles different genes play in different types of cancer, which could enable new kinds of therapies.
- New Research Shows How AI Can Act as Mediators
- Facial Expressions Of Mice Analyze With Artificial Intelligence
- Shell Begins to Reskill Workers in Artificial Intelligence
- Anthony Macciola, Chief Innovation Officer at ABBYY – Interview Series
- Brain Implants and AI Model Used To Translate Thought Into Text