Understanding the context of free text clinical documentation has been a huge challenge for making progress in development clinical decision support applications. While a trained doctor can scan a patient’s medical records in a few seconds to form a clinical impression based on years of highly skilled training, no clinical medical application to date has proven equally capable. Human’s are quite capable of quickly determining the semantic context of our written languages but lack the ability to quickly digest this context from 1000’s to million’s of examples in a short period of time. Building semantic context into your data is key to building ML applications that bring value to medical decision making. Consider this simple example from a doctors note:
Jane Doe, at 46-year-old female with a history of vertigo, complains of…
It doesn’t take much for any reader to quickly understand that we have Female Patient, with an age attribute, and a clinical finding of vertigo and it’s not a stretch for a trained Doctor to digest her 4000+ word medical record in a few minutes and ascertain a probable set of diagnostic orders to drive to a probable diagnosis.
At MedNition, the first step to building clinical context into data starts with clinically aware NLP and a approach that can digest 10’s of thousands of similar patient records, understand the semantic meaning, and find the patterns of the most appropriate clinic pathways for the benefit an individual patient.
The medical industry has build a strong foundation for applications designed understanding the semantic meaning of free text medical records by creating a global standard for medical terminology call UMLS. (Link to UMLS US homepage). The National Institutes of Heal (NIH) uses this for their PubMed and MedLinePlus applications. This extensive database of standardized Medical terminology and semantic relationships will help build interoperability between the next generation of medical applications and bring improved outcomes for patients.