Medicine has always been at the forefront of innovation, as quickly embracing new technology brings many benefits to patient care and discovering new treatments.

It’s not surprising that machine learning –a subtype of artificial intelligence–is becoming more and more popular for health organisations and in medical research. Machine learning in healthcare has many benefits, as it can act like a doctor’s office admin, diagnostic assistant, and research aid. It allows for an early detection of diseases, prevention and even improved patient adherence to treatment.

Read on to find out what machine learning can actually do and our case study of a health monitoring app.

What can machine learning actually do for healthcare?

Machine learning can handle an impressive amount of work that improves staff productivity, patient care and aids medical research.

• Anomaly detection - identifying events that deviate from expected results. For example high blood pressure and arrhythmia can both be considered outside of the norm
• Classification - categorising diseases and symptoms. It’s especially important for diagnostics and research 
• Clustering - labelling and organising medical events in clusters. Research benefits the most from accessing clusters and identifying relevant patterns in the data 
• Generalisation - the machine learning model’s ability to make accurate predictions based on new medical data 
• Automation - automating critical tasks such as managing health records, which are very time consuming 
• Prediction - predicting future health events based on current data, such as predicting a stroke based on age, blood pressure, family history, etc. 
• Natural language understanding - a lot of medical data is unstructured data, such as clinical notes. Natural language processing can help extract critical information from this type of data, information which would be otherwise underutilised or not utilised at all

How is machine learning be applied to healthcare?

Machine learning has many applications in healthcare - and we believe that there will be many more as AI advances. Here are the most important that will surely become indispensable in the near future:

Automated health record handling

Electronic Health Records (EHR) can be organised automatically, eliminating the need for a stressful and repetitive task.

ML can help improve the management and analysis of electronic health records, making it easier for healthcare professionals to access relevant patient information and make informed decisions.

Detecting and preventing diseases

Machine learning algorithms can analyse large sets of data - this takes into account sensor data from IoT-powered devices, medical records, radiographic images, etc. This saves a lot of time since it can be used to quickly detect patterns–and is more advanced than a traditional data system. This leads to early detection of all kinds of diseases, improving prognosis and treatment outcomes.

This is also true for predicting outbreaks and other events such as readmission patterns for individual patients making it easier to allocate the right resources and anticipate patient needs.

Improving treatment plans

Treatment plans play a huge role in treatment and recovery. Machine learning systems can take indicators such as genetics, clinical data, environment, medication, etc. and come up with an optimised and personalised plan.

For example, some patients may be high risk, but also allergic to certain types of medication. Providers can use a solution to find the most appropriate medication that will minimise the negative side effects, thus enhancing patient compliance with treatment.

Image analysis

Probably one of the most important uses of machine learning has to do with medical images (MRIs, CT scans, etc.). Systems can detect even the tiniest of anomalies and find risk patterns with a high degree of accuracy.

As a consequence, it helps with the skills shortage in many areas that lack radiologists. It also helps providers quickly identify anomalies that they may have missed otherwise.

Discovering new drugs

Discovering new drugs is important, but it’s also very time-consuming and usually a trial-and-error process. Scientists have to sift through a lot of datasets to find the most promising compounds and combinations.

This is exactly where machine learning plays a huge role. It can take years of research only to discard a drug due to its inefficiency for a certain type of disease. However, that same combination can be a lifesaver for another disease – and this matching can be done faster through an algorithm.

Promoting treatment adherence

Patient compliance with treatment is a major predictor of treatment success. However, doctors can only do so much to guide and encourage each patient.

Machine learning can help with this problem by providing apps that act as a health coach, reminding patients to take their medication, improve their diet, giving them personalised health tips.

For example, we built a gamified platform that helps GPs set up challenges (get more sleep, drink more water), complete with levels and experience points. This proved to be a much better way to engage patients and make sure they adhere to their treatment plans.

Medical research and development

It’s not just the discovery of new drugs: medical research improves every aspect of patient care. Machine learning can assist in analysing millions of research papers and finding crucial insights.

This leads to advancements in research, as it can identify new topics to explore and to advancements in treatment plans.

More accurate clinical trials

Rigorous clinical trials are an absolute necessity. However, the process of finding appropriate candidates doesn’t have to be this complicated anymore. Machine learning algorithms can quickly scan large amounts of data and sort candidates based on health indicators. Then, they choose the best ones out of hundreds or even thousands of people.

At Qubiz, we have worked on a solution that processes unstructured medical data such as plain text from hospital discharge notes. The data was then transferred to Electronic Data Capture systems (EDCs), used in clinical trials, thus enhancing the probability of a more accurate trial result.

Monitoring health

As mentioned before, monitoring individual patients’ health can be challenging, but it’s very important to do so.

Machine learning helps doctors monitor health outcomes in real-time. For example, wearable devices can send notifications directly to the provider when the patients’ health changes rapidly, enabling timely intervention in emergencies

Qubiz case study: health monitoring app

Our client has many years of experience delivering R&D projects across Europe. We have helped them improve a health monitoring app that can take accurate blood pressure, pulse and breathing rate readings in 40 seconds. All you have to do is to use a tablet or smartphone.

In order to achieve better results, we improved their data collection process by adding multiple regions of interest. A region of interest is a part of the patients’ face from which pixels are being collected. This is then averaged and sent to the cloud for further analysis.

These machine learning algorithms use the remote Photoplethysmography (rPPG) signal collected from the device to derive the patient’s vital signs.

We also improved the accuracy of the vital signs to optimise the signal coming from the camera, ensuring stability and minimising extraneous noise.

In this way, any mobile device can become a monitoring tool and even save lives.

What are the challenges of using machine learning in healthcare?

While there are many benefits to machine learning, there are also some challenges that stand out.

1. Bias can happen

Healthcare should be inclusive and representative. However, we know that this doesn’t happen all the time. A machine learning app that cannot generalise (i.e. learn) from new sets of data may not be appropriate for finding the best treatment options for all people (based on gender, age, race, medical history, etc.). So it’s very important to be careful how we handle the data.

2. The quality of data used

To make better predictions, machine learning systems need good data. And cleaning and validating data is a process in itself. Since healthcare usually deals with large amounts of fragmented and unstructured data, it’s important to work with skilled data scientists who can tackle these issues appropriately.

3. The skills shortage

The skills shortage doesn’t just refer to medical professionals. To effectively use machine learning, hospitals and clinics need specialised business intelligence skills and different roles such as: Data engineer, Business analyst, Software architect. This brings us to our next point…

Meet us at Medica 2023

We build great software so you can fully focus on delivering great healthcare. With more than 15 years of experience and many healthcare projects, our experts are here to help you build the best solutions, whether they are for monitoring, telemedicine, patient management, hospital business intelligence, etc.

‍

You can meet us at Medica 2023, the world’s largest medical fair in Düsseldorf, Germany. Find us at Messe Düsseldorf, hall 15, booth 15H10. Let’s discuss your needs and plans for improving patient care.

‍