Is natural language processing a healthcare game changer?

NLP-in-lettered-blocks

How comfortable would you feel if your entire health profile was based on text that could fit into a field in a form filled out at a healthcare provider ? All the in-person consultations during doctors’ appointments, images, emails, voice recordings are documented but unable to be integrated into your full medical storyline.

The frightening part is that there is probably much more unstructured data about you in other irretrievable formats than there is in your electronic health record. Even if some of those records were scanned, the ability to “translate” them into searchable format is probably very limited.

The convergence of artificial intelligence, machine learning and Natural Language Processing (NLP) now give hope that a total view of the patient is more than a pipe dream. What’s interesting is that the consumer market for voice activated devices like Siri and Alexa have provided a better understanding among the masses of what NLP really is. Yet we all know the frustration when we ask Alexa to “play some Beach Boys songs” comes back with “I can’t find any songs by the Beach Boys”.  Think of the frustration if a request for a endocrinologist’s commentary on your blood-sugar reports came up with the same response.

The challenge is even greater in healthcare NLP because it can literally be a matter of life and death, unlike reviewing one’s social media conversational patterns.

Add to this the fact that there is a very diverse population within the US healthcare provider and consumer universe with language and syntactic subtleties that challenge even the best artificial intelligence algorithm. As I wrote in a previous blog, even verbalizing the level of pain can have serious cultural implications that go beyond artificial intelligence.

The world class NLP platforms resolve these ambiguities and then convert them into normalized and searchable digital formats that can be shared across enterprises and more importantly with the patient/consumer.

One of the more interesting aspects of NLP is its ability to determine mood, emotion or sentiment in large bodies of unstructured or conversational text. I’ve had the ability to use a such an algorithm for content analysis and found it to be, as we say in Boston, wicked smart !

These NLP algorithms not only have the ability to identify “mainstream emotions” like fear, sadness, anxiety and doubt, but they are also able to unearth emotional patterns that have no clear definition. The data scientist told me that there was a certain “feeling” in the document that did not meet formal sentiment taxonomies, and that it would have to be given a new name.  I saw this as analogous to describing the undefinable taste sense called Umami, like describing the taste of a truffle to a friend.

From a budgetary point of view the accuracy of NLP searches across large volumes of information is directly related to the investment made in the technology. There is no room for error that create false positives (or false negatives) in healthcare segments like oncology. Think of this as being similar to investment in language translation platforms. It starts with the risk of translation using a free link to Google Translate and move upward where syntax and context can be embedded into the algorithm,

Many of the complaints related to some of these technologies have also been related to the composition and geographies of patient samples. One wild card for NLP is the social determinants that patients bring to the algorithm and how the unstructured data about that patient feeds into the logic.   Some of the early launches in AI and NLP received pushback because clinicians felt that the universe was biased to a certain institution that gathered the volumes of data and built the algorithm about it.

Finally, we cannot ignore the coevolution of AI, machine learning and NLP.  As more and more IoT devices on the edge (see previous blog) feed more reliable data into the AI engine, the ability to extract and validate more meaningful and health insights with NLP will improve dramatically.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: