Amazon Comprehend Medical adds relationship extraction to medical condition

Posted on: Jul 1, 2020

Amazon Comprehend Medical, is a HIPAA-eligible machine learning service that allows developers to process unstructured medical text and identify information such as patient diagnosis, treatments, dosages, symptoms and signs, and more. One of the more powerful features of Amazon Comprehend Medical is the ability to tie these entities together through a process called relationship extraction.  

We currently provide this feature for medication and test, treatments and procedures and are excited to announce that we have extended it to medical condition. If there is a mention of “left chest pain”, pain will be extracted as the medical condition with chest as the anatomical location and left as the direction. Previously some of these attributes would be unmapped but now with relationship extraction they are tied back to the relevant entity. For customers this means that they can index this information to differentiate between more complex medical conditions. For example if a customer is looking to do a survival comparison of patients who have non small cell lung cancer (NSCLC) based on the left or right lung, they can now quickly index and classify their patients. They can go further and view the predominant treatments to determine which options are best for their patient population.  

To learn more about Amazon Comprehend Medical, visit our product page or read our technical documentation.