Inside Our Tech: Machine learning basics and how we're using it to create a better experience for you
What Is Machine Learning?
According to Google’s Cloud AI Adventures, machine learning “brings the promise of deriving meaning from data.” Simply put, it's an automated system that uses data to answer questions.
Once enough sufficient data is collected, like how often you listen to Wu Tang Clan on Spotify, machine learning algorithms can be trained to perform certain tasks, like curating a weekly personal Spotify playlist consisting mostly of old-school hiphop. We experience some of these tasks everyday, like when Facebook recognises ours’ or someone else’s face in a picture when tagging it, or when Netflix recommends us new shows and movies similar to what we just finished binging.
Essentially, whether you’re using Netflix, Spotify or our EndoMetrix app, the machine learning technology that these services use collects relevant data about you to create a better, more personalised experience for you. The quality and quantity of data collected while using these programs determines how well they will be able to predict what you would like to watch, listen to or try as a treatment option next.
How Is Machine Learning Used in Healthcare?
Although the healthcare sector can sometimes seem like a late adopter in “cutting-edge” technologies like artificial intelligence (AI) - my mom loves to complain about the hospital computer software she’s been using since she started as a nurse in 1997 - machine learning has played an integral role in healthcare for at least the last five-to-ten years.
Since early 2013, IBM’s Watson - a data-learning AI tool - has been used in healthcare for managing clinical operations, building treatment awareness plans, collecting patient data to create individualised care plans and providing timely and targeted care to high-risk patients.
Other common uses of machine learning in healthcare today include image diagnostic tools to detect tumors throughout the body, drug development, radiology and radiotherapy, diabetic retinopathy diagnosis and robotic surgery for hair transplantation.
How Does EndoMetrix Use Machine Learning?
We want to help women by providing them with a personalised experience. We want to do this by using machine learning algorithms to help link each individual user’s symptoms to a specific condition like endometriosis, polycystic ovary syndrome (PCOS), adenomyosis, fibroids or vaginitis.
We are currently working on identifying patterns among our users by teaching our machine learning models that certain symptoms - like headaches, abdominal pain and excessive bleeding - can most likely be linked to a specific condition.
We are also teaching our model to learn how each user manages these symptoms. Whether our user relies on a specific diet regime, exercise routine or medication, her methods will enable us to suggest care options for users who have similar symptoms. This shared data will set our new users on an effective path to combat her symptoms with methods that both work for her and that will potentially work for others.
How Does Machine Learning Benefit Our Users?
Our app ultimately gives our users control over their own health and wellness. At her fingertips, the user will have access to a wealth of information compiled from both her own experiences and the experiences of a community of girls and women overcoming similar challenges. This data will help her curate a wellness plan covering diet, fitness, medication and meditation methods that targets her individual needs.
Not only will our users have access to this data goldmine, they will be in control over how their own data is shared with their healthcare providers and our general user base. Although our machine learning tools correlate this data with other users’ symptoms, each individual user’s identity will remain anonymous, and she can choose when and with whom she wants to share her data with publicly.
Although our machine learning model works to create a personalised and secured experience for each individual user within our app, these user experiences wouldn’t be possible without the community of girls and women contributing their own updated and accurate wellness journeys. Not only do our users help us help themselves, they help themselves help each other.
What are the Challenges of Working With Machine Learning?
According to Digital Ocean, the main issue with machine learning as it’s used more in more in healthcare and business are the “uncaught biases [that] perpetuate systemic issues.” These issues prevent people from opportunities like loan qualification, being shown ads for high-paying job opportunities, or receiving same-day delivery options.
Also, a recent peer-reviewed study has unfortunately shown that some machine learning programs exhibit human-like biases that include race and gender prejudices. For example, when using historical photographs of scientists while training a machine learning model, the model may not properly classify scientists who are also people of color or women.
The main challenge we face right now is gathering a sufficient amount of diverse data about our users to ensure our final product correlates their symptoms to a potential diagnosis and wellness plan accurately.
We just completed the creation of a survey to properly measure the symptoms, forms of treatments, diagnosis and support journeys of our potential users. We need at least 1,000 potential users of different ages, backgrounds and care experiences to complete the survey before we can successfully begin testing and eventually launching EndoMetrix.
What is the Future of Machine Learning in Healthcare?
In 2018, about one in five U.S. consumers have already used healthcare services powered by AI. Also, many of these consumers are open to using AI-operated clinical services like home-based diagnostics and virtual health assistants.
Within the next three years, the AI healthcare market is projected to become a $6.6 billion industry - up from $600 million in 2014. New machine learning technologies are predicted to include future applications like personalised prescriptions based on a patient’s genetic makeup and background, automated treatment - think of diabetic patients having their blood glucose levels checked and their insulin injected automatically using a small pump device - and autonomous robotic surgery for procedures like hip replacements.
All of these technologies are dependent on legal and regulatory questions that remain unanswered, including developer and product oversight, remaining transparent with the public and data privacy. Ultimately, these technologies have the capacity to completely alter the way we receive care and disrupt our traditional approaches to healthcare operations and delivery.
What is the Future of Machine Learning at EndoMetrix?
Although we don’t plan on implementing robots or DNA-influenced treatment plans into our technology makeup anytime soon, we do hope to run a machine learning model that one day encompasses millions of girls’ and women’s experiences across the globe.
The more diverse our user bases’ data is, the better we will be able to provide fast, effective and adequate care for them. Also, we hope to create a clear line of communication between our users and their healthcare providers, ensuring that our users are heard, taken seriously and believed.
We ultimately want our app to serve as a stepping stone to shortening the typical seven to ten years it takes to receive a proper diagnosis for gynaecological disorders like endometriosis. Eventually, this type of adaptive and individualised technology can be used to provide fast and adequate solutions for other therapeutic areas as well.
In order to achieve this future, we need your help now. It only takes eight minutes to complete our survey. The information you provide us today can help us build and provide an unbiased, accurate product for you and the rest of our community of girls and women who strive to receive a better wellness plan for their gynaecological challenges tomorrow.