To an untrained eye, data can look like a lot of 1’s and 0’s with no discernible pattern. To an experienced observer, the data turns into a picture. It shows us things we’d have never realized before its compilation. Machine learning is able to look even deeper and help turn a picture into a Picasso.
The idea is extremely potent and has wide ranging application. Take, for example, an article in TechCrunch that uses games as an example of the power behind the potential.
“Last year, Google’s artificial intelligence platform, AlphaGo, deployed techniques in deep learning to beat South Korea Grand Master Lee Sedol in the immensely complex game of Go, which has more moves than there are stars in the universe. Those same techniques of machine learning and AI can be brought to bear in the massive scientific puzzle of cancer.”
This same idea could be leveraged to help solve medicine's ultimate puzzle: The cure for cancer. Unfortunately, data needs to be readily available for machine learning to really do its thing. While decades of data exists regarding cancer treatment and results, much of it is unavailable because of one of two reasons. The most obvious is personal data security. HIPAA, the Health Insurance Portability and Accountability Act of 1996, guarantees user privacy and data security regarding personal information. It’s important to be sure, but, more often than not, pundits seem to agree that it’s become outdated.
The other issue involves a lack of innovation in healthcare. This is something we’ve spoken about often here and has cost the industry billions of dollars. This, albeit, is partially why HIPAA is so outdated. The world has changed since 1996 and technological advancement is a major reason why.
Consider this: HIPAA was finalized 11 years before smartphones were even a thing. Today we have social media, where personal information can be shared at the push of a button in 140 characters or less, and now offer SaaS based systems in the Cloud. Yet, even today, you have to provide HIPAA direction for every single provider you meet with, and it can limit your care. Need to visit a new doctor? Better sign some paperwork. Need to be reviewed for a second opinion? Better resign the same paperwork. It’s a system of inefficiency, not only for patients but for researchers as well. TechCrunch touched on this fact briefly in their article.
“Many data sets, including medical records, genetic tests and mammograms, for example, are locked up and out of reach of our best scientific minds and our best learning algorithms.”
Now, we aren’t advocating for publishing everyone’s personal information and data for the purpose of research, far from it in fact. However, there would seem to be a way to provide data access without sacrificing an individual’s privacy. In 2015, there was even a Bill that made it through Congress that would have paved the way for such opportunities. The bill was ultimately reworked to remove the provisions that would have altered HIPAA laws, but the conversation was a start. Just because the right answer has yet to be found, doesn’t mean that it eventually won’t. There is some good news in all of this. Even without flushed out and updated regulations, some programs have the flexibility to provide some data to get the research kick started. TechCrunch discussed a few.
“A number of large-scale, government-led sequencing initiatives are moving forward. Those include the U.S. Department of Veteran Affairs’ Million Veteran Program; the 100,000 Genomes Project in the U.K.; and the NIH’s The Cancer Genome Atlas, which holds data from more than 11,000 patients and is open to researchers everywhere to analyze via the cloud. According to a recent study, as many as 2 billion human genomes could be sequenced by 2025.”
While the information currently available lacks the breadth that exists behind locked cabinet drawers it still shows promise. The ultimate eventuality of data sharing and machine learning could lead to early recognition of illnesses and accelerate the development of new drugs for things like cancer. If the opinion of TechCrunch is an accurate one, then we’re three steps from seeing it all happen.
- It should be easy for patients to share their data including records, imaging and testing results. It would be easy and legal for this kind of sharing to occur if a common consent form was mandated and in place in the field.
- There’s a lot of funding in the medical world, but more is needed when it comes to technology and machine learning and data science. “Just as the Chan Zuckerberg Foundation is funding new tool development for medicine, new AI techniques need to be funded for medical applications.”
- Data and research needs to be wider reaching across gender and racial diversity. According to a research study by the University of California San Francisco, clinical studies still miss nearly 40 percent of the U.S. population. If we achieve results in machine learning, it is important that they are applicable to all.
Once the items above are in place, machine learning can do its job and work to reach the maximum of its potential to paint its proverbial Picasso. Like his art, even if the average patients can’t decipher the data, they’ll still be able to appreciate the results.