NATURAL LANGUAGE PROCESSING: A Brief Tutorial for Health Systems

Key Concepts in Natural Language Processing


Presented by Frank Manion, PhD, FAMIA


Dr. Manion is VP for Innovation at Melax Technologies where he guides development of next-generation NLP applications. He has held c-level positions in biomedical informatics and IT at two major comprehensive cancer centers. His own research is in the area of formal semantics and biomedical ontologies.


Dr. Manion presented the webinar, Natural Language Processing: A Brief Tutorial for Health Systems for the Houston TX chapter of HIMSS in February, 2021.


This eBook provides an overview of key concepts presented during Dr. Manion's webinar.


What is Natural Language Processing?


Natural Language Processing is a branch of Artificial Intelligence and Computer Science that applies algorithmic approaches to derive data and information from textual information.


Why does it Matter?


Studies have shown that 80% to 90% of available information and data in Electronic Health Record (EHR) text artifacts (notes, H&P, radiology, pathology, etc.) remain accessible ONLY through the text notes.



Is NLP Being Used in Healthcare Settings?


Yes, there are both public domain and commercial software packages available for clinical NLP. Systems are rated across a number of tasks to achieve a F1-performance of at least 90% accuracy. National and international challenge projects are used to test NLP systems against known results to check accuracy level.


NLP is highly regarded by:

  • The National Institutes of Health

  • The National Cancer Institute

  • The Centers for Medicare & Medicaid Services

  • The pharmaceutical, health IT, insurance and academic medicine industries


National Committee for Quality Assurance (NCQA.org)


"Moving to more uniform implementation of national electronic data transfer standards and use of technology such as natural language processing are some of the ways we can begin to address this [quality measures/HEDIS].”


How can clinicians and researchers access the 80% to 90% of data available only in text notes?


Natural language processing's methodology identifies relevant unstructured data and relational information.


Extracting data based on relationships makes the information actionable. For example, in a patient's medical history, NLP can find and exclude familial history, so that a family member's health history is not confused with the patient's. NLP is able to determine whether the patient has had a history of hypertension, or if hypertension is a family trait.


Similarly, NLP can verify whether a symptom noted in a chart relates to a specific condition, such as COVID.


Examples of NLP uses


NLP Learns Quickly


so Users see Results Quickly


Machine learning-based methodologies identify and learn patterns and features from existing annotated text. Once the patterns are identified, maching learning-based NLP can quickly employ that learning to other unstructured data.


Rule-based NLP learns from a pre-determined set of codes and rules and recognizes those patterns in documents. It's a good system for one-off projects.


Melax NLP Builds Pipelines to Facilitate Information Flow


With a pipeline, a user can amass information on any number of conditions, co-morbidities or cohorts within a set of unstructured data. Melax offers 13 pre-built pipelines that help track cancer, social determinants, and COVID-19 signs and symptoms, to name a few. Melax also provides 12 additional sub-pipelines to help users take a deeper dive into clinical data.


Pre-built data extraction conduits provide seamless use, while also offering the capability to apply advanced Melax technology if needed.


The Melax Technologies NLP product line:

  • Melax's flagship tool - CLAMP, Clinical Language Annotation Modeling and Processing toolkit

  • LANN our literate annotation tool

  • VITAL for EHR chart review

What our Clients say...


Ciitizen's goal is to empower seven billion citizens across the planet with all of their health data. LANN gives us the ability to rapidly scale our annotation efforts across our clinical NLP models by providing a comprehensive and easy-to-use interface both for the data scientists and subject matter experts. Ciitizen was impressed by Melax Tech’s hands-on team of bioinformaticians and the support we received to achieve our goals. We value Melax Tech's cooperation in building the next generation of toolsets.

Tom Corcoran, Data Scientist, Machine Learning Engineer



Melax Tech | https://melaxtech.com | contact@melaxtech.com

713.208.8195