I imagine a world in which AI is going to make us work more productively, live longer, and have cleaner energy. –Fei-Fei Li, Professor of Computer Science at Stanford University

Artificial intelligence has the potential to improve every aspect of our lives and help us transform healthcare. Let’s have a look at how healthcare is practiced today and how AI is transforming it.

Healthcare implies keeping the health of an individual up to the mark or improving it. It covers injuries as small as paper cuts to blood cancer.

Healthcare can be divided into three categories, namely the following.

  • Curing
  • Preventive
  • Predictive

We can use the huge amount of data produced every day to find a better cure for a disease, find new drugs, and even predict the probability of a disease long before any symptoms relating to it are observed.

Healthcare industry problems

The problems of the healthcare industry can be divided into two broad categories. One category of the problem arises from the sociopolitical and financial issues, while the other arises from the technological challenges in the industry. Issues like shortage of beds, shortage of healthcare workers, and unqualified medical practitioners belong to the first category. The second category contains issues like slow research, human errors in analyzing data, and the lack of data transparency among the organizations.

We’ll be focusing only on the technological challenges in this post.

AI to improve healthcare

Artificial Intelligence

Artificial Intelligence offers an amazing opportunity to transform the world in a huge manner. It has been called as the new electricity by Andrew Ng. It has the potential to touch every person’s life in a meaningful way, just like electricity did.

In healthcare, AI can help in improving each step of the ecosystem. From the prediction of disease to finding a new drug to making all new gene modifications.

Let’s look at what the potential holds for the future.

AI-Healthcare ecosystem

Imagine a scenario where a couple is about to get married. An AI system can check the compatibility of their genes to figure out if there is any risk to the child or some gene that can result in a complication in the child’s normal life. This system can then help in figuring out the right measures that can be taken before and after the baby is born.

Suppose the system identified a problem with a particular gene, we could then alter this gene to remove its harmful effect. The AI can also help in discovering the right drug that could help in keeping the problem in check even after the child is born.

The child was born healthy and is now a teenager; she is wearing a health tracker like Fitbit, which keeps track of all her vitals like the heartbeat, steps taken in a day and, calories burnt in a day. These readings are used by her AI assistant to tell her about the changes she needs to make in her routine to continue her healthy lifestyle.

Unfortunately, one day she is in an emergency and is being taken to the hospital. Her Fitbit reading could be sent to the paramedics to take decisions even before they arrive at her place. The AI system can tell the possible issues that she might be suffering from, like cardiac arrest, etc.

The blood sample that is taken while on the road can be easily analyzed by a computer vision system to make the preliminary diagnosis. Currently, most of the diagnosis is manually made by an expert by looking into the microscope and studying the cells.

After her release from the hospital, the past data analyzed by the AI system will predict the probability of her readmission to the hospital and will suggest the appropriate measures to prevent it. This can be done through constant reminders of following the drug dose. Intelligent medicines can also be prepared that sends a signal when it has been taken by the patient to truly make things automatic.

With increasing age, her AI assistant will continuously collect the data to predict health and will take appropriate preventive measures to keep her health to the best possible level.

This important lifelong data will be used by the system to improve itself and make things much better from the next instant.

AI in action

Digital Diagnostics using Computer Vision

Currently, a lot of diagnostics require a trained professional to analyze samples of blood, saliva, tissues, semen, etc. under a microscope. This is very time consuming and error-prone. Dedicated machines exist for different tests, but a cheaper solution is possible using AI.

Digital diagnostics use computer vision technology to analyze images of these samples and then apply algorithms such as ANN and CNN to figure out the size shape and movement of cells. This data is then used as the features to train a machine learning model to find the problems that the patient might have.

Similar technology is also being used to analyze X-Rays and CT Scans. Convolutional Neural Networks are very good at analyzing images. They use filters to find features of the image, which is not possible using the normal feature engineering techniques.

Predicting Spread of Virus Outbreaks

Various machine learning models have been used to predict the spread of viruses and other infectious diseases. Social media data from platforms like Facebook, Twitter, etc. are used to fit regression models to predict areas of next outbreaks.

Patient flow optimization

We can use data like the number of patients per hour visiting the hospital, current weather conditions, and common injuries to predict the number of patients that might come to the hospital on a given day. This intelligence is useful for medical centers to optimize their supplies and be better prepared for emergencies.

 Personal Doctors

Advances in Natural Language Processing has made it possible to create smarter chatbots to help patients at any hour of the day. A user can simply type in the common symptoms that she is facing, and her chatbot will tell her if she should see a doctor or not. The assistant can also book an appointment with the doctor automatically based on the urgency of the situation.

NLP helps in finding the “intent” of the user from the sentence that the user has typed. Techniques like stemming and lemmatization, stopword removal are used to preprocess the data. This preprocessed data is then fed into models like LSTM to figure out the intent of the person and then accordingly find a response to it.

24×7 Monitoring

When a patient is under observation, doctors and nurses need to do regular visits to keep track of the vitals of the patient. This takes up a lot of time and also leads to emergencies due to the intervals between the visits. AI systems are now capable of tracking this data all the time and predict if something wrong is going to happen. Timely alerts generated by these systems are helping save time as well as lives.

Time-series forecasting is one of the methods used in such a system as the data received is a stream of values with time. Recurring Neural Networks can also be used to analyze such data as RNNs are good at predicting future values based on the past values in a stream.

Challenges

The AI-Healthcare ecosystem described above though very idealistic, is already happening currently but is not as connected as it should be. Here are some of the challenges that the current industry faces.

  • The IoT of healthcare is not very easy to implement. The data lives in silos; a Fitbit can’t communicate with the hospital system; the digital pathology can’t communicate with the other system in the hospital. If the patient’s health records are from a different hospital, then that data can’t be taken by the new hospital as currently, the data is kept by each organization privately.
  • No standards exist around processing, storing, privacy, and sharing of the healthcare data. Every organization follows the standards laid out by their IT team or vendor. All this makes the data very difficult to share between the organizations and systems. Policies of national and international levels are needed to bring this ecosystem together.

Ethics in Healthcare

Ethics is one of the most important pieces of the puzzles when we are talking about AI in healthcare. I leave it to the reader to think about the following scenarios and realize how complex it could get when we have intelligent machines making decisions for us.

  • Who owns your data? The Electronic Health Record(EHR) that your hospital has belongs to you, but should you be allowed to take ownership of it? What if you had a very rare disease and your data is of prime importance, should the society be allowed to use the data even though you don’t want it?
  • Suppose the AI system finds out that you are very likely to have a type of cancer that is incurable. Would you like to learn about it? Think about the emotional toll it can have on the person.
  • What if the predictions made by AI were wrong. Who should be responsible for that, is it the developer who coded it or the organizations that made the system or the data that was used to make the system in the first place?

AI in healthcare has a huge potential if we can solve some of the aforementioned issues. We see tremendous advancements in the area, and most of the things described in this article are not as fictional as they sound.