Medical Pure Language Processing Tech Has Come of Age

Medical Natural Language Processing Tech Has Come of Age
Dr. Tim O’Connell, Founder & CEO of emtelligent

For a few years, pure language processing (NLP) has held the promise of dramatically growing the flexibility of healthcare organizations to shortly and precisely perceive unstructured medical textual content in medical notes. Utilizing medical NLP, healthcare suppliers, medical researchers, and payers would uncover significant insights hidden in unstructured textual content sooner, with fewer errors, and at much less price than guide knowledge overview and evaluation. This high-quality medical-grade knowledge in flip would drive advances in understanding illness development, assessing therapy efficacy, and detecting inhabitants well being traits and different use circumstances.

Issues haven’t fairly labored out that method.  Whereas single-institution, single-document-type NLP tasks have confirmed viable, coping with the complexity of language throughout a number of establishments and doc sorts has eluded correct NLP. 

One mistake healthcare organizations generally make is that they assume the medical NLP software program they bought is enough for their use case. But these instruments merely should not correct sufficient to supply clinical-quality NLP as a result of they don’t seem to be fluent within the language of drugs. Medical vernacular is filled with inherent complexities akin to important ambiguity, a particular lexicon, and heavy use of localized medical shorthand. Add within the variety of medical specialties and dearth of requirements for the construction of medical paperwork, and it’s clear that healthcare organizations require extremely specialised medical NLP that leverages superior applied sciences akin to synthetic intelligence (AI) and deep studying.

Thankfully, AI fashions have improved drastically with the appearance of deep studying. Nonetheless, with none form of medical experience guiding the event of those deep studying fashions, customers find yourself with outcomes that principally say, “We discovered an entire bunch of issues. Now you go work out what’s vital.” That’s not precisely clinical-grade data.

What’s wanted for a high quality medical NLP platform is a mix of expertise and medical experience. By infusing deep studying fashions with specialised medical experience, trendy medical NLP software program may help suppliers, payers, pharmaceutical firms, and medical researchers get essentially the most worth from the info.

Limits of conventional NLP in drugs

Whereas NLP undoubtedly has proved helpful to researchers, the method concerned could be labor-intensive and time-consuming. Let’s say researchers wished to make use of NLP to seek out all sufferers in a goal inhabitants who had appendicitis final month, with the info for use in a white paper. Low-precision conventional NLP could determine 1,000 sufferers – however the NLP could be steadily flawed.   In consequence, a researcher should undergo the info and ensure all of it. Granted, that’s nonetheless higher than the researcher laboring over guide chart opinions, however that also falls properly in need of an environment friendly and efficient resolution. 

For different healthcare use circumstances – akin to understanding human speech, computer-assisted coding (CAC), and medical determination help – even “principally correct” software program is nowhere near acceptable. Healthcare organizations that implement a medical NLP platform that’s extraordinarily correct will have the ability to apply their knowledge to makes use of circumstances past analysis.

Selecting the best medical-grade NLP platform could be tough for healthcare organizations that will not know exactly what options or performance will work for them. Listed here are three issues to search for in a clinical-grade NLP platform:


Does the platform present sufficient accuracy in your group’s functions? For instance, some platforms could have an algorithm for negation detection, the method of figuring out the presence or absence of situations or ailments akin to most cancers or diabetes. Nevertheless, the accuracy of those algorithms can range relying on their capacity to contextualize language in medical notes. 

The power of an NLP platform to precisely determine frequent medical phrases, together with slang, have to be a precedence to ensure excessive ranges of accuracy. Annotation software program can carry out the work of physicians who historically would annotate 1000’s of medical stories – and do it a lot sooner – however the medical NLP resolution should have the ability to maintain tempo in velocity and accuracy. 


Healthcare organizations should know the precise functionalities of a medical NLP platform. Which ontologies does it help (SNOMED, RadLex, MEDCIN, ICD-10, and so on.)? Does it determine questions or uncertainty? Can it extract insights from the unstructured textual content of medical, diagnostic, and semi-structured stories?

One other vital characteristic entails the platform’s capacity to determine relations inside knowledge. Does it determine measurements? Or dates? If the platform is analyzing a report a couple of affected person with a historical past of appendicitis, does the algorithm perceive that appendicitis occurred previously? Or does it simply say that the affected person has appendicitis now?  If the report accommodates a press release that the affected person’s mom has breast most cancers, does it attribute breast most cancers to the affected person, or can it precisely determine the expertise?

Deployment location

Many medical NLP distributors provide solely cloud-based providers, however not all healthcare organizations are wanting to ship their affected person knowledge to the cloud. Right this moment’s give attention to data security makes cloud-based options on this house much less enticing. For these organizations, on-premises medical NLP platform deployments are important. 


Relative to the necessities of supplier and payer organizations, medical NLP for too lengthy has left a lot to be desired in accuracy and suppleness. Current advances in AI now make it attainable for medical NLP to assist healthcare organizations leverage extremely correct knowledge for medical work, analysis and drug growth. Healthcare organizations ought to guarantee a medical NLP platform is correct sufficient and contains the options they should get essentially the most from their knowledge.

About Dr. Tim O’Connell 

Dr. Tim O’Connell is the founder and CEO of emtelligent, a Vancouver-based medical NLP expertise resolution. He’s additionally a training radiologist, and the vice-chair of medical informatics on the College of British Columbia.

Open chat
Thank you for contacting us, more info please chat.