The basics of AI and Health care
Overview of the many different applications of AI in Healthcare
Before writing this, I just used Chat-gpt to give me the simple definition of AI (since I didn’t really have any knowledge beforehand). Very soon I also have to start and finish an assignment using Chat-GPT for my accounting assignment (Don’t worry it is in the instructions to use it to search for accounting frauds!). Thinking back on the past couple of months we have had so many different developments in AI that I can barely list them.
Even in our daily lives, we use AI on a daily basis. Here are a few of the MANY examples of AI used.
- Virtual assistants — like as Siri, Google Assistant, and Alexa
- AI algorithms for personalized recommendations — used through youtube, Instagram, Spotify, and even Google search
- Natural language processing systems — such as Google translate
- Navigation systems providing real-time traffic updates
However, the biggest growth and arguably the most important field out of all of the fields that are growing is the use of ai in healthcare.
This article will talk about the basics of each growing use of ai in healthcare.
Basics of AI
A simple way to think of Artificial Intelligence (or shortened to AI) is basically simulating human intelligence in machines by allowing them to learn, problem-solve and make decisions.
The way you can think of machines before this ability is small children who cannot understand algebra. However, once they 1) learn it from their school teacher, 2) do the homework 3) review it once again. Once they do all of that they are now able to learn how to solve problems they haven’t learned before + the same ones on tests and assignments.
In the field of AI, there are many types of problems it can be used to solve. There are 4 key types of AI, all linked to the memory they possess. Here are them:
Reactive Machines
- this type has no memory whatsoever and has very limited capacity. These machines cannot use their gained “experiences” to know what to do presently — they basically cannot learn.
- These types of machines are used mostly to perform simple tasks such as Spam filters or Netflix recommendations
Limited Memory
- These machines have memory and are able to learn from their history. Nearly all current AI systems fall under this category.
- Some of the many examples are chatbots and virtual assistants to self-driving vehicles are all driven by limited memory AI and image recognition
Theory of Mind + Self-aware
- Both of these types of AI only exist hypothetically. Both are able to be more human, doing human things such as understanding. Self-aware AI is one step above the theory of mind and will possibly be basically completely self-aware (I am not that sure we should get to this point since that would also mean the end of humanity as we know it.. possibly some sci-fi movie.. )
As for the different types of AI, there are three types.
- AI — the broader field, includes lower-class models unable to learn by themselves
- Machine Learning is a subfield that is broadly defined as the capability of a machine to imitate intelligent human behavior.
- Deep Learning; an artificial intelligence function that imitates the workings of the human brain in processing data and creating patterns for use in decision making
Key Terms:
- Neural Network
A neural network is an artificial attempt to simulate the network of neurons that make up a human brain so that the computer will be able to learn and make decisions in a human-like way.
Firstly there is input data, then it goes through hidden layers, and finally, outputs based on the output layer. Each layer consists of any number of neurons (or the bubbles in the diagram) which each provide information to the model. Here are the steps:
- The inputs/information from the outside world is provided to the model to learn and come to conclusions from. Input nodes then pass it on to the hidden layers.
- There is any number of hidden layers in a neural network — it only depends on the type of neural network it is.
- From the different criteria provided the model finally outputs the ‘answer’. This all depends on the type of answer you want — if it's binary then there will be one answer, but if it is a multi-class classification then there will be more than one.
As mentioned there are many types of neural networks out there. Here are the key different types of neural networks.
- BNN; lighter weight, but sometimes lower accuracy neural networks
- CNN; is a class of deep neural networks, most commonly applied to analyzing visual imagery
- RNN; is a class of artificial neural networks where connections between nodes form a directed graph along a sequence, making them especially useful for sequential data
- DNN- Deep Neural Network; a neural network with multiple layers between the input and output layers, DNNs are especially
well suited for dealing with unlabeled or unstructured data
2. Types of Learning
There are three key types of AI learning; supervised, unsupervised, and reinforcement.
- Supervised Learning; a machine learning task of learning a function that maps an input to an output based on an example input-output pairs
- Reinforcement Learning; dynamic programming that trains algorithms using a system of reward and punishment.
- Unsupervised Learning; is a branch of machine learning that learns from test data that has not been labeled, classified, or categorized. Instead of responding to feedback, unsupervised learning identifies commonalities in the data and reacts based on the presence or absence of such commonalities in each new piece of data
Within each of these types, there are a few key types of AI used which will be learned more about later in the article.
Now onto the tech used in the medical industry:
AI in Healthcare
Healthcare refers to the improvement of a patient's health through the diagnosis, treatment, or cure of disease, illness, injury, and other physical and mental impairments in people. This is helped by a variety of healthcare professionals.
One of the biggest problems in Canadian healthcare is the long wait times and limited access to physicians and medical technology. AI is able to aid in making the process all the faster for patients and more efficient.
There are many different applications of this technology in healthcare, but here are some of the main ways that will be talked about in this article.
- Medical image analysis
- Personalized treatment plans
- Drug discovery
Medical image analysis:
Medical image analysis is a relatively new but promising way to diagnose and treat a variety of medical conditions. This is done with the aid of machine learning algorithms and advances in medical imaging technology, being able to analyze large amounts of images with both accuracy and efficiency.
One of the biggest applications of medical image analysis is in the field of radiology. For many years, radiologists have used their expertise and experience to interpret X-rays, MRIs, and CT scans to identify various medical conditions. However, this process is time-consuming and may lead to errors in the diagnosis. With medical image analysis, algorithms can be trained to recognize patterns in medical images that show a variety of different medical conditions. Now radiologists can use this and be able to diagnose conditions much faster.
This technology is especially important in critical care situations such as identifying cancer in patients, where time is of the essence. In addition, medical image analysis can help reduce the risk of errors in diagnosis, which can have serious consequences for patients.
However, there are still some downsides to this technology. A big risk is overfitting — where the model fits exactly against its training data. This can lead to false positives and false negatives in diagnosis, which can have serious consequences for patients. Another problem is that this analysis requires large amounts of high-quality medical images to be trained effectively.
Personalized treatment plans:
This section of AI coincides with the field of precision medicine — one that I have been exploring for the past couple of months. If you would like to learn more about this field, I would suggest reading this article I wrote a couple of weeks ago:
Traditionally, medical professionals have relied on a one-size-fits-all approach to treatment, often prescribing the same medication or procedure to every patient with a particular condition. However, every person is unique, and their medical needs are no exception. One medicine that could “cure” a condition, could be poison to another.
AI could help in this by identifying patterns and making recommendations for personalized treatment plans. This means that patients can receive treatments that are tailored to their specific needs, rather than a one-size-fits-all approach.
In a nutshell, this is how this will work.
- Medical professionals will input data into the algorithm, including the patient’s medical history, genetic information, and lifestyle habits. The algorithm then analyzes this data and identifies patterns that are relevant to the patient’s condition
- A personalized treatment plan that is tailored to their specific needs will be created as per the algorithm.
- Finally, the patient will be able to make informed decisions based on what the algorithm recommends. This has the potential to significantly improve patient outcomes and reduce the risk of adverse reactions or side effects.
However, this field is not really that big — more so an idea. There are some ethical issues with this such as the privacy and security of this data as well as the accuracy and the negative effects it could have if the algorithm is wrong.
Drug discovery
The pharmaceutical industry is a necessary component of the healthcare system, involved in the research, development, manufacturing, and marketing of drugs and medical devices.
Along with the formulation of drugs, the pharmaceutical industry also looks for cost-effective drugs that can also be easily absorbed by the body. Consistent proteomic testing can allow the manufacturers to stay in competition with others while still having high-quality drugs that meet regulatory standards.
However, the development of the drugs to go into the market is a long process that can take up to 20 years to develop. These technologies make this process easier and aid in shortening this process for the companies and making it easier. Further, when the companies can prove the drug to be safe, it would take less time to verify it to the regulators such as the FDA or HPFB.
There are many different technologies that researchers can use to identify different drugs and potential drug targets (such as proteomics, genomics, etc. ), however, this process is very long. This process can become detrimental if there is a global pandemic — such as covid-19 — and we need a vaccine to treat people ASAP.
The solution to this is using AI algorithms to analyze vast amounts of data and identify potential new drug candidates for a variety of diseases.
Hey! Thanks for reading my article! It means a lot to me!!
My name is Anya Ishani Sharma, and I am a grade 11 student passionate about personalized medicine and tech! If you found this review interesting and would like to discuss it with me please feel free to reach out. You can contact me through my email (anya.ishani.sharma@gmail.com), and my Linkedin (click here).
