AI-Powered Health Services
Welcome to the beautiful world of artificial intelligence!
Artificial Intelligence! Many were unfamiliar with this word at the beginning of the 21st century, but it has now become established in vocabularies. As the name suggests, this branch of computer science focuses on creating an “artificial” form of thinking similar to human thinking yet even faster and more accurate. In the realm of healthcare, AI is currently being used to mimic doctors and medical researchers in disease diagnosis as well as analysis and presentation of medical data.
So how is this different from traditional forms of technology?
Well, what distinguishes AI is its ability to gather, process, and derive a conclusion from the data. In other words, the need for human intervention in AI-powered services is minimized. This is achieved through AI’s two powerful tools: machine learning algorithms and deep learning.
What is Machine Learning?
The International Business Machines Corporation, otherwise famously known as IBM, defines machine learning as a technology that “focuses on applications that learn from experience and improve their decision-making or predictive accuracy over time.” In short, what machine learning algorithms do is: learn from data, identify patterns, make decisions. Sounds familiar? That is actually precisely how we — humans — operate. We learn from our experiences (walking during childhood, exams during school, interactions during adulthood, etc.). Then, we use those experiences to make decisions in our daily lives. For example, when you experience that a particular food tastes “good,” it is very likely that you will choose it over a less-favorable food whenever given a choice. Or, when you experience that a specific behavior puts you in danger, you will most likely try to avoid that action.
Machine learning encompasses several strategies, usually broken down into three categories: supervised learning, unsupervised learning, and reinforcement learning.
In supervised learning, a machine is given example inputs and their corresponding outputs. Subsequently, these act as a “teacher” for the algorithm, hence enabling it to “learn” a general rule that can be used to map the inputs to outputs. Now, whenever given an input, the machine will be able to predict the outcome based on the general rule that it learned earlier.
Unsupervised learning is different. In this strategy, only inputs are given to the algorithm, without any labels — i.e., a set of numbers are given without knowing whether they represent a car’s value or a house’s value. Hence, the algorithm is left to find structure in the inputs.
Last but not least, in reinforcement learning, an algorithm is given a certain goal in an interactive environment. By trial and error and using provided feedback (rewards and punishments for positive and negative behaviors), the program learns from its own actions and experiences.
What is Deep Learning?
First, to understand the relationship between Deep Learning, Machine Learning, and Artificial Intelligence, we can use a Russian dolls analogy; deep learning sits inside machine learning, which sits under artificial intelligence.
In deep learning, an algorithm is given raw inputs, meaning that we don’t specify which features are significant in the data. Instead, the algorithm decides this for itself. In other words, deep learning is essentially the component of AI that aspires to act like the human brain!
An essential aspect of deep learning is neural networks. These are inspired by the biological neurons found in our brains. Neural networks are used to recognize relationships between large amounts of data. A “neuron” — commonly referred to as a node— in these networks is a mathematical function that collects and classifies the data based on a specific architecture. However, it is not just one level of calculations. Neural networks usually have many levels between the input and the output, called layers. Nodes (or neurons) in each layer may be connected to different nodes in layers below or above. Ultimately, these nodes move the data through the network, from one layer to the next. The more layers an algorithm has, the deeper it is called — hence the name “deep learning”!
How are these AI-driven Technologies aiding Healthcare?
Among its vast potential in the field of healthcare, Artificial Intelligence can be especially beneficial for disease diagnosis and treatment recommendations. Indeed, there has been a significant focus on AI-powered diagnosis and treatment techniques since the early 1970s, when MYCIN was developed at Stanford University.
MYCIN
MYCIN was an early AI-driven expert system that was used to diagnose blood-borne bacterial infections. It also recommended antibiotic treatments for the infection, with the dosage determined based on the patient’s body weight. Since many antibiotics have the suffix “-mycin”, the program’s name is derived from these treatments.
MYCIN is considered a simple algorithm when compared to today’s machines. It operated using a knowledge base of approximately 600 rules. The physician would have to answer a series of yes/no or textual questions and, in the end, MYCIN would create a list of possible bacteria that caused the infection, ordered from high to low based on the probability of each diagnosis. The recommended treatment was also listed in front of each bacteria.
Watson
IBM’s Watson Health unit is currently one of the forerunners of AI-based healthcare. One of the general goals of Watson is to aid medical personnel in the treatment of patients. Unlike MYCIN, Watson is not a single algorithm but rather a set of “cognitive services” connected through Application Programming Interfaces — or APIs for short. These services include — but are not limited to — speech and language, vision, and machine learning-based data-analysis programs.
When using Watson, first the physician must pose a query to the system, describing the patient’s symptoms and other related information. Watson then analyzes the input to determine the most critical pieces of information. Besides, the algorithm also delves into patient data to discover other facts relevant to the patient’s medical and hereditary history. Examining the available data and forming and testing hypotheses, the algorithm provides a list of “individualized, confidence-scored” recommendations in the end.
Others
Over the past few years, more and more tech firms and startups have set a goal to aid AI applications in healthcare. Google, for instance, is collaborating with health delivery networks, like the UK National Health Services, to develop and implement prediction models from big data to warn clinicians of high-risk conditions, such as heart failure and sepsis.
There are also several companies and organizations that focus specifically on the diagnosis and treatment for certain diseases — including cancer — based on patient genetic profiles. Such firms can be especially significant considering the complexity of understanding all genetic variants of a disease for physicians without the aid of AI. Foundation Medicine and Flatiron Health, two firms owned by Roche, specialize in this approach.
“Population Health” machine learning models are also becoming essential for predicting populations at risk of particular diseases or outbreaks. Since many factors contribute to outbreaks, including history, environment, and even socio-economic status, the powerful data analysis techniques that these models utilize make them effective at making predictions.
From data analysis to making diagnoses and giving prescriptions, AI-powered services have been tremendously aiding healthcare — especially in the past couple of decades. But, there is still a long way for them to become completely independent of human interference. It is still — in most cases — physicians that input the data in the machine learning algorithms. More importantly, AI algorithms still operate based on the scientific and medical knowledge that humans feed them. Still, AI’s beneficial applications in healthcare by making predictions and analyzing medical data — and ultimately saving lives — are undeniable. And let us not forget how AI has gotten to where it is today: by mimicking our very own beloved, all-powerful brains!
About That Medic Network
If you like this article, be sure to follow us for more content on our Website, Facebook, Twitter, Instagram, LinkedIn.
Feeling friendly or excited for our next article? Email us at info@thatmedicnetwork.com or subscribe to our newsletter.
References
- “Artificial Intelligence (AI).” IBM, 3 Jun. 2020. https://www.ibm.com/cloud/learn/what-is-artificial-intelligence. Accessed 26 Mar. 2021.
- Brownlee, Jason. “What is Deep Learning?” Machine Learning Mastery, 16 Aug. 2019. https://machinelearningmastery.com/what-is-deep-learning/. Accessed 28 Mar. 2021.
- Chen, James. “Neural Network.” Investopedia, 23 Dec. 2020. https://www.investopedia.com/terms/n/neuralnetwork.asp. Accessed 28 Mar. 2021.
- Davenport, Thomas, and Ravi Kalakota. “The potential for artificial intelligence in healthcare.” Future healthcare journal vol. 6,2 (2019): 94–98. Accessed 28 Mar. 2021.
- “Machine Learning: What is it and why does it matter?” Suit of Analytics Software (SAS), https://www.sas.com/en_ca/insights/analytics/machine-learning.html#:~:text=Machine%20learning%20is%20a%20method,decisions%20with%20minimal%20human%20intervention. Accessed 26 Mar. 2021.
- Osiński, Błażej, and Budek, Konrad. “What is reinforcement learning? The complete guide.” DeepSense.ai, 5 Jul. 2018. https://deepsense.ai/what-is-reinforcement-learning-the-complete-guide/. Accessed 28 Mar. 2021.