Machine learning in healthcare: solving challenges in innovative ways

16.10.2019
7 min.

The healthcare industry is rich with data coming from patients, medical devices, and the environment. The more data you collect, the more you can learn and discover. However, analyzing all this data manually is an unachievable quest. The goal of machine learning consulting is to put this data in perspective and deliver valuable insights that are actionable for humans. The ability to process and benefit from such data allows tailoring medical care to every patient, relieving medical staff from tasks that can be automated, and predicting outbreaks.

Offering personalized care

Machine learning algorithms contribute to patients’ health by offering personalized treatment and 24/7 support at a lower cost compared to medical staff’s billable working time.

Personalized medicine

Nowadays, when receiving patients, doctors are limited in their knowledge of patients’ conditions. They can only examine current symptoms, excerpts from the medical history, and restricted genetic information. Hence, different patients are likely to receive identical prescriptions for similar symptoms. For example, if they have a urinary tract infection, they will most likely get Bactrim.

With insights gained from machine learning though, each prescription could accommodate a patient’s medical history, genetics, past conditions, diet, stress levels, etc. Having this information will make it possible to personalize the treatment rather than go with the standard benchmark. For example, it would be possible to create a unique variation of particular medication to avoid side effects which a particular patient is likely to develop.

Behavior modification

Many of the prevalent diseases are manageable or even avoidable. Often, type 2 diabetes, obesity, and heart diseases can be avoided by practicing a healthier lifestyle. However, adjusting the lifestyle requires a change of behavior, which is not a one-time effort but the one requiring constant reminders and follow-ups. For this purpose, machine learning algorithms can aggregate data incoming from patients’ connected devices and sensors to generate insights on these patients’ behavior and guide them in this transformational journey. 

Altering smoking habits with SmokeBeat  

SmokeBeat is an innovative application that passively gathers user data on smoking behavior. The application uses an accelerometer on a smartwatch or a smart band to detect hand-to-mouth gestures. SmokeBeat processes this data and offers real-time Cognitive Behavior Therapy incentives. User responses to those incentives are constantly measured and recorded to improve effectiveness. Additionally, SmokeBeat compares users’ smoking data with their peers of choice, creating a sort of a supporting social network.

SmokeBeat offers information on smoking patterns

Encouraging people to change behavior

There are three ways in which machine learning-based apps can influence user behavior.

The first one is by reaching out to users during appropriate moments. Healthcare apps will send alerts for different purposes, for example, to remind of medicine intake. However, if the timing of such alerts is wrong, it will not only be discouraging but even harmful. Receiving alerts while driving can increase the odds of a car accident. On the contrary, user-aware apps can understand user behavior and send alerts at the optimal moment.

Another way of influencing user behavior is by automatically logging users’ activities, which can make them more self-aware and prompt behavior changes. This category includes applications that offer a comprehensive view of user activities such as sleep patterns, exercising, time spent in a car, etc. 

Finally, it is possible to prompt behavior changes by sharing relevant health resources with users. Very few people are interested in general-purpose newsletters. Customized healthcare resources are more appealing. For example, given that a user is a workaholic, the healthcare app can inform on how overworking can shatter glucose balance.

Virtual nursing

In today’s busy hospital environment, nurses can get overworked, struggling to offer enough personalized support to every patient. Healthcare facilities count on virtual nurses to solve this issue. Virtual nurses are computer-generated avatars who can interact with patients like humans. They are designed to be social, empathic, and informative.

Virtual nurses can interact with patients more regularly than human nurses, and answer questions in-between doctor visits. They offer quick answers (faster than waiting for a nurse), and they are available 24/7.

One example of a virtual nurse is Molly. This is a female avatar able to remotely monitor medical conditions, which would be challenging to monitor on the spot. Molly receives data such as blood pressure and weight from monitoring devices connected via Bluetooth. These devices are positioned at patients’ homes, which makes it convenient to take measurements as often as needed.  Additionally, Molly understands human speech and can respond in kind. It also offers a chatbot for private discussions.

Molly, the virtual nurse services the elderly

Alleviating doctors’ pressure

In addition to benefiting patients, machine learning is valuable for the medical staff. It can speed up diagnostics, increase precision, and minimize surgical penetration.

Diagnosis and disease prediction

People constantly die in hospitals due to preventable conditions such as sepsis just because the warning signs are not recognized in time.

Machine learning can significantly improve this situation as it catches details that a human eye can miss. It shortens the time between medical imaging and diagnosing, and it can validate diagnoses.

Medical imaging

Even with all the technological advancements, medical image analysis is a tedious task prone to human error since it requires great attention to detail. Machine learning algorithms can detect even the subtlest changes in medical scans. Furthermore, traditional scan analyses (such as CAT scan and MRI) are time-consuming. To shorten this cycle, a MIT-led research team has developed a machine learning solution that can analyze medical scans 1,000 times faster than doctors.

Disease prediction

By analyzing large sets of data, machine learning can find patterns and predict diseases. The University of Nottingham developed an algorithm able to predict heart attacks. Their machine learning-based system gathers patients’ medical data such as blood pressure, smoking habits, age, etc. and predicts which patients are likely to have a stroke within the next ten years.

Another application was successful in predicting suicide risks among patients admitted to a particular hospital. The system gathers and analyzes hospital admission data such as medication history, previous diagnosis, age, address, etc. The results are used to predict suicide attempts. At the trial stage, the system successfully predicted 84% of suicide attempts among 5,000 patients.

One concern about machine learning applications is that it is not always possible to explain how they arrive at particular predictions. This “black box” nature is a challenge when it comes to healthcare. Doctors don’t feel comfortable making life-death decisions based on something they cannot explain.

Robot-assisted surgery

Using robots in healthcare is not new. Robotic assistance in surgery increases precision, allows reaching different areas of the human body with minimal penetration, and alleviates pressure from human surgeons as robots can take over some parts of the work. Accenture estimates that by 2026 robot-assisted surgery will have saved the US healthcare industry $40 billion.

Doctors using robots to minimize harm

The Da Vinci robotic surgical system is becoming increasingly popular in operating rooms. Da Vinci allows surgeons to operate robotic limbs while performing surgeries, which adds precision and reduces tremor. It also allows making incisions as tiny as 1-2 cm. Additionally, it offers a high-definition view of the surgical site. 

Performing surgery using the Da Vinci surgical system

HeartLander is another example of a surgical assistant. This mini‑robot enters the chest through a tiny incision. HeartLander can perform sensing, mapping, and treatment across the heart surface.

Heartlander robot enters the heart from a tiny incision

Automatic suturing

Medical researchers have long attempted to automate suturing as it reduces the length of surgical operations and takes some pressure off surgeons. One such attempt is the Raven robot designed specifically for laparoscopic (abdominal) operations. When tested, the Raven robot sutured with an 87% accuracy. This is ever more important considering that abdominal suturing is a complicated procedure that discourages surgeons from performing laparoscopy.

Using innovative research methods

Using machine learning makes innovative research initiatives possible in healthcare. This includes predicting outbreaks and preparing for them in advance. It also becomes possible to speed up the conventionally long drug discovery process, let alone using datasets uploaded by ordinary people for medical research purposes.

Drug discovery

Drug discovery is an expensive and long process. Thousands of elements need to be tested together, and only one of them might end up as a viable drug.

Machine learning algorithms are used in the drug discovery process for the following purposes:

  • Minimizing clinical trial duration by predicting how potential drugs will perform
  • Identifying combinations of existing drugs which can form a new treatment
  • Discovering new drugs based on compound testing
  • Finding new uses for previously tested substances

IBM and Pfizer collaborate on oncological research

Pfizer, a pharmaceutical company, is using IBM Watson for its immune-oncology research. While a human researcher can read around 300 articles a year, Watson was able to process one million journal articles and data on four million patents. Given the ability of machine learning systems to ingest a huge amount of information, they can identify non-obvious connections and help create treatment plans out of drug combinations.

Identifying health patterns using public datasets

In another research project, machine learning algorithms were employed to study medical data of 90-years-olds to identify patterns in characteristics that keep this demographic healthy. Those patterns can later be used to develop drugs disabling harmful genes.

Outbreak prediction

Nowadays, a huge amount of data can be collected from satellites. This includes real-time data from social media and other historical web data. Machine learning algorithms help aggregate this data and make predictions about potential disease outbreaks. One example is predicting malaria outbreaks by analyzing data including monthly rainfall, temperature, and similar parameters.

This can be particularly relevant in third-world countries that lack medical infrastructure and necessary education to combat those diseases. Knowing about such critical outbreaks upfront will allow taking necessary precautions to minimize their negative impact and save lives.

ProMED reporting disease outbreak online

ProMED (the Program for Monitoring Emerging Diseases) offers an online real-time reporting system showing outbreaks of infectious diseases worldwide, and any exposure to toxins affecting human or animal health. ProMED aggregates data from sources such as official reports, media reports, local observers, and reports contributed by its subscribers. An expert team reviews these reports before they are accepted into the system.

The data provided by ProMED is aggregated by HealthMap to visualize disease outbreak in every country.

Map rendered by HealthMap displays disease alerts worldwide

Crowdsourcing data

Crowdsourcing allows researchers to aggregate large amounts of data uploaded by people from their mobile devices and healthcare platforms and using it with their consent. With the advancement of the internet of things, there is even more data available for experimentation. Healthcare organizations are attempting to put this data to its best use to understand diseases and make changes to treatments.

Publicly available health datasets help in understanding cancer  

It is familiar for people to upload and label pictures. However, it is also possible to upload medical datasets. If those datasets are labeled and formatted properly, they can be processed by machine learning algorithms, which can lead to a better understanding of particular medical conditions or contribute to the development of new drugs.

One such example is the CancerBase initiative. It allows patients with cancer to share their health datasets anonymously. This data can be used by medical researchers to understand and treat cancer. 

[Patients] can donate information for the greater good. In return, we make a simple promise: when you post data, we’ll anonymize it and make it available to anyone on Earth in one second. We plan to display this information like real-time traffic data. HIPAA doesn’t apply to this direct data-sharing.

Jan Liphardt

Takeaways

Machine learning in healthcare provides opportunities to use large amounts of data that cannot be processed with human efforts. Incorporating insights from this data into healthcare practice brings many benefits:

  • Reducing costs, such as of drug development
  • Reducing time: machine learning is faster in diagnosing than human doctors
  • Saving lives: predicting disease outbreaks allows preparing in advance, which increases the chances of survival
  • Improving patient satisfaction: virtual nurses are available 24/7 to answer patients’ requests. Additionally, insights derived from aggregated data make it possible to customize treatment for each and every patient 

Finally, a cautionary note. Despite all the advantages of machine learning , it is still important to be wary of the ethical and moral issues that come with it. Machine learning algorithms embody the values that humans are coding into them, and for now, the accountability for this is not regulated. Additionally, some healthcare practitioners fear being replaced by algorithms in the future, which makes them unwilling to use machine learning in their practice.

Tags: