One of the main things we do at M51 is automate on a technology basis and make entrepreneur ideas scalable on a global basis. Technology is an important part of our operations and we keep a watchful eye on its many development. In that regard, artificial intelligence is one of those buzzwords that you hear all the time. It’s everywhere, and its persuasive power to transform numerous industries is slowly catching up with healthcare.
This relatively newfound focus has produced numerous pilot programs, studies, research partnerships, and dozens of startups looking to understand the instant value of AI and come up with mainstream AI algorithms with lasting impact on patient care.
Healthcare is a particularly interesting subject because it’s closely tied to the greater good – who doesn’t want technology to improve our health and maybe save some lives? So, in that positively-thinking spirit, here is a quick look at the impact AI is having in healthcare.
Voice-based detection and prevention
Artificial intelligence has a key role when it comes to detecting, preventing, and even treating various diseases, both common and uncommon variety. In that regard, voice technology is making a lot of noise (pun intended). In terms of interface, nothing beats conversation when it comes to communicating intuitively so it’s no surprise voice AI is aiming at healthcare for further development and implementation.
At the forefront are natural language processing (NLP) methods that extract information from unstructured data sets (e.g. notes) to augment structured medical data. The brunt of AI-powered voice assistance revolves around providing a personalized experience. This includes low-level research for relevant information such as hospitals in the area with the least waiting times, selecting a personal doctor, and managing and maintaining the health routine, but also using voice tech as triage and seeking treatment tips for specific illnesses.
The more advanced side (and arguably cooler) is predicting and diagnosing conditions based on voice. Based on the patient’s speech pattern, tone, and pitch, voice AI can be used as an indicator for something as simple as the common cold to severe neurological and degenerative diseases.
audEERING, an intelligent audio processing and paralinguistic analysis company, is an example of technology capable of identifying early symptoms of Parkinson’s disease by analyzing subtle changes in voice. If you’re wondering how capable technology really is, it reportedly achieved 92% classification accuracy for healthy and diseased samples.
Through a verbal cue, AI can detect subtle problems with mood or mental abilities and prevent abnormal physical or emotional conditions (Amazon is already working on a patent). This is especially important with early onset detection and estimation of disease progression as vital parameters to deliver timely and effective treatment of more serious diseases.
Voice AI will continue to generate a demand for the use of health technologies that employ real-time data to engage with patients in a more personalized way. That will place digital healthcare, accessible from smart devices, on top of the trend list as people will want a streamlined and hassle-free medical experience: quick response and no visits or routine check-ups.
Voice tech also shows a lot of promise in reducing the barriers in patient engagement, where access to more sensitive or subjective information can be further eased. Some companies claim readmissions are reduced by lowering infection risk thanks to the usage of voice assistants for online visits and remote monitoring.
I’d argue there’s even an emotional side to consider here, as patients just might be more open to communicating to a machine rather than being vulnerable in front of a doctor. This both makes and doesn’t make sense but then again – we humans are strange beings.
Machine learning-based in-depth analysis
The mention of AI in healthcare is immediately turned to the implementation of complex algorithms in the analysis of complicated medical data. Perhaps it’s because the value of mimicking human cognitive functions is best evident on the hardest challenges. Maybe it’s because of the potential to simplify complex and time-consuming medical activities.
In any case, thanks to machine learning (ML), much of possible applications of AI are moving out of the realm of theoretical and having a practical effect in treating patients, specifically when it comes to doing more with less. Thanks to ML techniques that examine structured data such as imaging and genetic data, AI is inferring information from deep data analysis and leading to improved clinical decision making.
In fact, in my ‘Companies that have mastered business intelligence’ post, I mentioned Aidoc. That is a fine example of AI involvement in healthcare and an indication of how far technology can progress for the good of humankind. In a nutshell, what Aidoc does is employ AI solution that detects high-level visual abnormalities from different types of medical scans (CT scans, X-rays, etc.). The aim is to provide assistance and improve human diagnostic power by flagging the most acute and urgent cases where a faster diagnosis and treatment could mean the difference between life and death.
One other example is the fight against breast cancer. Deep learning, a facet of ML that utilizes a layered algorithmic architecture for data analysis, can correctly predict the chance of a woman developing breast cancer. Such is a deep learning model created by a team from MIT and Massachusetts General Hospital (MGH) that can deduce from a mammogram if a patient is likely to develop breast cancer as much as five years in the future. This model proved to be significantly more precise at predicting risk than traditional models, accurately placing 31% of all cancer patients in its highest-risk category as opposed to only 18% for existing models.
These are just some instances that have proven AI’s worth in targeted and controlled applications, becoming advanced enough to provide assistance in a meaningful manner, particularly for diagnostic purposes like radiology, pathology, and others.
Cool case in point – Empatica’s Embrace
Wearables such as smartwatches and fitness trackers were among the first devices to embrace the health angle so it shouldn’t be all that surprising that there is a device that monitors and alerts for the most dangerous kinds of seizures. Specifically, Embrace is a smartwatch for epilepsy management powered by advanced machine learning. It identifies convulsive seizures and sends alerts to caregivers through the sensor of motion and physiological signals, while also providing standard health tracking activities like sleep, rest, and physical exercise.
One underestimated aspect of this device is the way it collects data – passively because it’s hard for users to be aware of exact data when the seizure happens so there are not many actions that are required of them. Empatica, the company behind Embrace, achieved an accuracy rate of 98% during the clinical testing and has received clearance from the FDA for both adults and children, while previously being approved in Europe as a medical device for seizure monitoring and alert.
From diagnostics to smart devices, artificial intelligence has become a nurse to a human doctor in many ways. A large part of it is thanks to technology’s ability to learn and adjust without human intervention so it can work on its own.
While the examples mentioned in this post are already showing results and lots of promise, it’s going to take at least a few years for AI to fully reach its potential, if even then. Right now, AI is filling gaps and being an important part of a conversation that wasn’t available before. Everyday we get closer to patient-friendly care that is safer, more accurate and consistent. Sometimes, that’s all you can ask for.