Explainable A.I. Or Why You Need To Understand Machine Learning In Healthcare

A doctor in China uses a machine learning algorithm to detect signs of pneumonia associated with SARS-CoV-2 infections on images from lung CT scans. Epidemiologists in Canada are using the technology to monitor the spread of a disease and help prevent outbreaks. In the U.S., researchers are using artificial intelligence for more efficient drug discovery . Elsewhere around the world, patients are turning to their phones to access symptom checkers leveraging smart algorithms.

These instances where medical professionals and patients alike employ artificial intelligence (A. I.) are already happening but in the coming years will be even more commonplace. However, as the technology becomes ubiquitous at a heightened pace, understanding how it works and explaining its deductions becomes increasingly challenging.?

However, if the future of medicine and healthcare relies on a collaboration with A.I. , we will have to be able to understand the underlying processes of these tools so that we can in turn trust their insights. Seeking such transparency is what explainable A.I. deals with.

In this article, we will go over the basics of this disruptive technology; as well as highlight the importance of better understanding it whether you are a healthcare practitioner or a patient.

The need for explainable A.I.

Explainability when it comes to A.I. refers to humans understanding the output of an algorithm, in particular a machine learning (ML) one. Often, the latter is considered as a black box that not even the programmers behind such models can fully understand or explain how they achieved a certain result.?

And when such software programs are dealing with data as sensitive as healthcare-related ones, we can appreciate the need to better understand how they arrived at a specific result. Such a grasp on the technology will allow us to manage it, gauge its efficiency and eventually trust it. Moreover, this will help address eventual challenges and concerns arising from A.I. in medical practice.

As such, the aim with explainable A.I. (XAI) is to shift the traditional black-box approach to a white-box one for greater transparency, interpretability, and explainability. These are the three core elements that are often highlighted in this context. While the quest for XAI in life science is relatively new, efforts are underway to produce “glass box” models.

Researchers call it crucial for those using A.I. in medicine to understand what it is. They even go as far as saying that “omitting explainability in clinical decision support systems poses a threat to core ethical values in medicine and may have detrimental consequences for individual and public health”. We can go further by saying that not only clinicians but also patients, and indeed any stakeholder in healthcare, should better understand medical A.I. This is because in the digital health age, A.I. will become a necessary tool for every player in this field. In the next section, we will go through its basics to help you get acquainted with the common terms and how they work.

Algorithms, artificial intelligence and machine learning?

In understanding A.I., it’s important to make the distinction between simple algorithms and artificial intelligence . Let’s consider each of their definitions separately.

Algorithm

An algorithm can be defined as “a step-by-step procedure for solving a problem or accomplishing some end”; and while such instructions are part of what makes an A.I., they aren’t defined as such by themselves.?

Artificial intelligence

A.I. on the other hand is defined as “the capability of a machine to imitate intelligent human behavior”. However, when talking about A.I. nowadays, we most often focus on machine learning (ML), which is one of A.I.’s subcategories.?

Machine learning

ML is itself defined as “the process by which a computer is able to improve its own performance (as in analysing image files) by continuously incorporating new data into an existing statistical model”.?

With ML techniques , developers enable algorithms to learn a task without being explicitly programmed for this particular task. Such ML algorithms can identify patterns in datasets given good quality and quantity of data.?

While there are several ML subtypes, below we elaborate on the three major subtypes as well as an advanced method, deep learning (DL), that are more relevant to healthcare. We will use a child and a teacher analogy since algorithms can also be seen as children learning new skills through their developer teachers.

1. Supervised learning

This ML subtype is comparable to teaching a child exactly what to learn. It is used when the exact task of an algorithm can be precisely defined with the data at hand. In medical practice, it can look as follows. We have two patient groups, Group A and Group B, each with their own set of medical records. Group A’s set contains the family history, lab markers and other details of the diagnosis. Group B’s set consists of the same types of information, but the diagnosis is missing. We can train an algorithm with supervised learning to assign the right diagnosis to Group B, based on the associations and labels the algorithm learns about in Group A. This method is the most frequently used training mode.

2. Unsupervised learning

As the name suggests, this method is akin to learning without a teacher. The starting tools are there, but the child decides on the end result. We provide different datasets to the algorithm and it finds associations on its own, even those we might not have thought about. Additionally, we do not modify the algorithm based on the outcome. Such a model can discover new drug-drug interactions or cluster patients according to the attributes they display.

3. Reinforcement learning

Reinforcement learning shares similar features with unsupervised learning in that the starting tools are given to the “child” and it is left to make decisions on its own in order to achieve a task. However, unlike unsupervised learning, reinforcement learning involves input from the “teacher”. After a series of actions (but not after each action as with supervised learning), A. I. developers input their feedback to nudge the algorithm towards the best course of action. The issue with using this subtype in healthcare is that we cannot test out the algorithm on a large number of scenarios since patient lives are at stake.

4. Deep learning

DL is an advanced subtype of ML that holds different potentials altogether. Its functioning is based on artificial neural networks (ANN), which are themselves inspired by the neural network of the human brain. DL consists of a layered ANN structure where the more layers it has, the more complex tasks it can perform. Let’s say we are building a model to group patients based on their diagnosis.?

If the information reads “Type 1 Diabetes”, an ML algorithm will cluster medical records with “Type 1 Diabetes”. A DL algorithm on the other hand will be able to, with time, assign patients with only the “T1D” abbreviation mentioned in their records to the same group, without human input. Other ML subtypes will require manual input from the developers to recognise this abbreviation.

Know your A.I.

While understanding artificial intelligence and machine learning algorithms at a deeper level might require advanced programming knowledge, it will become important for non-programmers to understand the basics of the technology, especially in healthcare. This was the aim of this article and to help guide you further through the intricacies of A.I. in healthcare, we have more resources to share.

To help you draw the line between simple algorithms and A.I., we wrote an article dedicated to the importance of this distinction in medicine. Last year, The Medical Futurist Institute published a paper in npj Digital Medicine to guide medical professionals in better understanding the basics of A.I. and its potentials in medicine.

For an even deeper dive into the world of A. I. in healthcare, we published an e-book on the topic earlier this year. ‘A Guide To Artificial Intelligence In Healthcare ’ is aimed at being a comprehensive guide to help readers get a firm grasp of the possibilities and limits of A.I. in healthcare.

However, the world of A.I. and its implications for healthcare are ever-evolving and require close scrutiny. To that effect, we will be sharing regular explainers and relevant analyses on The Medical Futurist.

--

Dr. Bertalan Mesko, PhD is?The Medical Futurist ?and Director of?The Medical Futurist Institute ?analyzing how science fiction technologies can become reality in medicine and healthcare. As a geek physician with a PhD in genomics, he is a keynote speaker and an Amazon Top 100 author.

Get access to exclusive content and analyses about?the future of digital health on Patreon.com !