AI Able to Predict the Level of Recovery Following a Stroke

By

    

AI Able to Predict the Level of Recovery Following a Stroke


Background information:

What is a Neural Network?

What is Logistic Regression?

What is Random Forest?


_________________________________________________________________________________

        In this article: AI to Predict Rehab Outcomes in Stroke Patients, computer scientists have developed a software program using a Deep Neural Network (DNN) model, logistic regression (LR), and random forest (RF) with the ability to predict motor function in stroke patients 6 months following a stroke. Although it is a primary research article, I do recommend that you give it a read as it provides a good primary example of how AI is being designed to serve the medical field.

    Scientists analyzed the data of 1,056 stroke patients aged 43-72 between January 2003 and January 2020 without other health complications. Using Python, TensorFlow 2.3.0, and sci-kit-learn toolkit 0.23.2, these machine learning (ML) algorithms were able to detect patterns from this input data using 14 different variables like age, sex, type of stroke, and more. The deep neural network, logistic regression, and random forest were able to accurately predict upper limb recovery with a 91%, 87%, and 88% accuracy respectively. For lower limb recovery, the algorithms were able to predict outcomes with an 82% (DNN), 76% (LR), and 80%(RF) accuracy. As a reference, 70-80% is considered to be an acceptable accuracy of prediction, 80-90% is considered to be excellent and above 90% is considered to be outstanding.

     A similar study (Predicting post-stroke activities of daily living through a machine learning-based approach on initiating rehabilitation - ScienceDirect) conducted in 2018 used random forest and logistic regression and were only able to predict rehabilitation outcomes with an 80% and 79% accuracy respectively. This means that the algorithm used to predict these outcomes are becoming increasingly refined and soon enough, we may be able to have these software programs predict outcomes with uncanny accuracy. 


   All in all, the accuracies of each of these predictive algorithms have been found to range from outstanding to excellent for upper limb recovery and excellent to acceptable for prediction of lower limb recovery. Being able to accurately predict motor abilities following a stroke allows clinicians to determine better rehabilitation programs and better counsel their patients. For example, if the program predicts that a patient will undergo a full recovery, doctors can confidently ease patients', as well as their family's, worries. Additionally, as algorithms are being continuously improved, one day they could serve to predict abnormalities occurring within the body better than human specialists which would save thousands of lives.


What do you guys think are potential drawbacks/implications to relying on AI to determine rehab programs?


What other ways could AI like this be used to potentially innovate the field of medicine?


What other problems could these algorithms potentially be used for?

Popular posts from this blog

Can robots make pizza?

By
Image
One of the most outstanding questions in the Artificial Intelligence field is how robots can help humans to automatize repetitive tasks and tedious work. We, as humans, love pizza, but we do not always have the time (nor the skills) to make one, so… can robots make (good) pizza for us? Well… researchers at MIT think so!   Pizza Dough   Nowadays, more and more companies and higher-education institutions are investing vast amounts of resources in developing Artificial Intelligence (AI) systems and robots. Companies such as Boston Dynamics, Kawasaki, Google, and Zebra have spent billions of dollars on developing cutting-edge robotic systems. Additionally, educational institutions such as MIT, the University of Pennsylvania, and the University of Tokyo are currently researching novel AI algorithms and techniques. Researchers are working on designing robots that can adapt to mutable environments and challenging situations. More and more engineers and computer science students are specializi
Read more

Can AI do your homework ?

By
Image
  Can AI do your homework ?   When we were young, all of us wished that one day we can have a robot that does our homework for us, with simple click, all of your assignments are done and you get to play video games all day. Well that's not a dream anymore, thanks to AI which refers to artificial intelligence. AI has reached a point where it is capable of doing some of the things that human beings can do. Advancements in technology are enabling computers to think, reason and answer questions like never before. Yet, most people still have reservations about whether or not AI can complete something as complex as homework assignments. The first step to completing any assignment is finding resources that will be able to answer your questions or help you get started on the work you're doing. Thanks to all the resources and libraries the internet has, it's more suited for AI to find solutions to homework problems that are not easy for human beings Nowadays, there exist AI that doe
Read more

Can AI become self-aware?

By
AI and Cognitive advanced mechanics are the two robust fields of plan and designing, holding onto all the spotlight recently. Man-made consciousness is a human knowledge recreation handled by machines, while mental mechanical technology is the end product of advanced mechanics and mental science that makes arrangements with mental peculiarities like getting the hang of, thinking, discernment, expectation, memory, and consideration. Mechanical technology is a piece of AI, where the robots are customized with man-made brainpower to play out the undertakings, and AI is the program or calculation that a robot utilizes to fill mental roles. In less difficult terms, a robot is a machine, and AI is the astuteness fuel strengthened to light perceptual capacities in a machine. There are various kinds of AI had been discovered:  Receptive Machines AI: Reactive machines are modified with an anticipated result in view of the information it gets. They neither record past experiences or capacities n
Read more