SPEAKER: Thanks to machine learning, we can extract knowledge from medical records, call center conversations, medical voice soundbites, medical forms, regulatory filings, research reports, insurance claims, pharmaceutical documentation, and more.
This ultimately helps doctors and care teams get holistic views of their patients quickly, or health plans to see population trends for their members, or pharma to draw insights from drug development research.
This is possible thanks to a field known as natural language processing, which is concerned with programming computers to process and analyze large bodies of human communication that can live in many different formats, such as written texts, spoken utterances, or even official documentation.
And so in this episode, we will share how organizations can use one of Googles natural language services to specifically help process structured and unstructured health care language data using NLP, which stands for Natural Language Processing.
[MUSIC PLAYING] The Healthcare Natural Language API contains four key features that help you find, assess, and link knowledge in your data in the following ways.
For one, it has the ability to text to medical concepts, which is referred to as knowledge extraction.
It also identifies and connects related medical attributes, which is known as relation extraction.
It can also assess surrounding factors that could be clinically relevant, known as context assessment, and standardizes medical concepts so they can be analyzed across systems, known as knowledge linking.
Another way of thinking about NLP is that it can extract critical clinical information, like medications, medical conditions, as well as understand contexts, like negation, such as, this patient does not have diabetes.
It also understands temporality, such as, this patient will start chemotherapy tomorrow, and even infer there are relationships between things, such as side effects or medication dosage.
And whats most notable is we have a long list of ontologies the natural language models are trained with.
Two notable ones are the ICD, which is used to code and classify morbidity data from inpatient and outpatient records, physician offices, and most National Center for Health Statistics surveys.
Theres also SNOMED clinical terms, which provides core terminology of electronic health records.
The models also include US official codes for insurance procedures and RxNorm, which contains a list of all medications available.
Technical practitioners can leverage healthcare NLP to build apps for their own organization or for their entire industry, such as enabling precise search and discoverability across patient populations, enable exchange of digital health care information and avoid processing delays, establish and maintain regulatory compliance, or automate administrative workflows, such as identifying and eliminating errors that require corrective activities.
Anyone who has science-based texts around research or clinical data can get value from our Healthcare NLP API.
Here are a few sectors that can benefit to name a few.
In telehealth, you now have off-the-shelf support for exchanging medical knowledge captured in written form and can extract structured knowledge from texts and make it available for digital services, such as chat bots, call centers, clinical decision support systems.
This also frees up time by triaging patient calls and resolving cases that do not require the intervention of a clinical professional.
Pharmaceutical researchers are also enabled via a standard patient discovery interface for population health and R&D applications, since papers and clinical trial documentation can be surfaced to match patients or find novel treatments.
Those managing clinical trials can both increase their number of participants, as well as process the high volume of feedback and decrease time to government approval.
Users who manage billing, especially in insurance companies, can have even better integration with claims payment or automate billing and coding for insurance.
You can enable the Healthcare NLP from your Google Cloud Projects UI or via the command line.
If you do not have a project, theres a link in the description along with other helpful resources.
And once you have setup permissions you can begin using its context-aware models to extract medical entities, relations, and contextual attributes.
Each text entity is extracted into a medical dictionary entry.
To extract medical, texts make a POST request and specify the following information in the request.
First, the name of the parent service, including the project ID and location.
Also, the target text.
The maximum size is 10,000 Unicode characters at this time.
Lets quickly show you how you can leverage the Natural Language Healthcare API.
Lets test out the sample medical record for a hospital patient.
We send over the sample medical tests to the API on the backend, and we render the JSON response in this web app.
In the first panel down here, we can see the various entities extracted and their corresponding medical code.
Next, we can see their diagnoses with their corresponding confidence scores.
Now looking at the relationships between the entities, we can actually group together important attributes.
For example, take a look at how long and how much of these medications are taken or were prescribed.
There are so many possibilities when using Healthcare NLP, and especially when you pair it with Google services, such as Dialogflow AI for a chatbot interface.
You can even build custom models to build low-code apps using AutoML Entity Extraction for Healthcare, so that users can simply upload documents and then perform manual annotations to train and build a model on what they would like to extract.
This can then be integrated into larger data pipelines.
You can also use Document AI to process faxed documents or enable enterprise search for life science organizations using the Google Knowledge Graph.
To learn more about the Healthcare Natural Language API, you can visit cloud.google.com /healthercare/do cs/concepts/nlp.
And community, if you found this episode helpful, please subscribe to the channel to get notifications of more health care episodes.