Amazon Comprehend Medical provides natural language processing, Medical Named Entity and Relation Extraction (NERe), Protected Health Information (PHI) Detection, ICD10CM ontology linking, RxNorm, and SNOMED CT ontology linking, enabling a broad range of applications that can analyze clinical text. All requests are measured in units of 100 characters (1 unit = 100 characters), with a 1 unit (100 character) minimum charge per request.
Natural language processing: Amazon Comprehend Medical Medical Named Entity and Relation Extraction (NERe) can be used to extract medical information from natural language clinical text quickly and accurately.
Protected Health Information (PHI): The PHI detection API finds and tags the 18 specified PHI fields according to Safe Harbor.
Ontology linking: The Amazon Comprehend Medical Ontology linking APIs, such as ICD10CM, RxNorm, and SNOMED CT, attempt to locate the relevant medical entities in the natural language clinical text and then infer which ontology code is most suitable to the detected entity. For example, ICD10CM will attempt to detect all medical conditions in the clinical text and then link the most appropriate ICD10CM code.
You can estimate your cost using the AWS Pricing Calculator.
Free Tier
Amazon Comprehend Medical offers a free tier covering 85k units of text (8.5M characters, or ~1000 5-page 1700-character per page documents) for the first month when you start using the service for any of the APIs
Amazon Comprehend Medical pricing
Amazon Comprehend Medical pricing examples
Example 1
You are working on a project that required medical named entities such as medical condition and medication to be tagged in the records for thousands of patients and want to run a pilot on 1000 medical charts to test Comprehend Medical’s performance. We assume you run this outside of the free tier using the Comprehend Medical NERe API and that each chart has 5 pages on average with 1700 characters per page.
Total charge calculation:
Total units: 1,000 charts * 5 pages per chart * 1,700 characters per page / 100 characters per unit = 85,000 units
Total cost: 85,000 units * $0.01 per unit for NERe = $850 for the pilot
Example 2
You compare the test charts against your ground truth tags and find the results promising and develop your solution. When you go into production you will process roughly 35,000 charts a month using the Comprehend Medical NERe API.
Total charge calculation:
Total units per month = 35,000 charts * 5 pages per chart * 1,700 characters per page / 100 characters per unit = 2,975,000 units per month
Total cost: 1,000,000 units * $ 0.01 per unit + 1,000,000 units * $0.005 per unit + 975,000 * $0.001 = $15,975 per month
Example 3
You do refine your pre-authorization model and find that you can improve performance if you use SNOMED CT concepts. You switch the workload in example 2 to use the SNOMED CT API which returns SNOMED CT concepts and will process 35,000 charts a month.
Total charge calculation:
Total units per month = 35,000 charts * 5 pages per chart * 1,700 characters per page / 100 characters per unit = 2,975,000 units per month
Total cost: 1,000,000 units * $ 0.0075 per unit + 1,000,000 units * $0.00375 per unit + 975,000 * $0.00075 = $11,981.25 per month
Additional pricing resources
Easily calculate your monthly costs with AWS
Contact AWS specialists to get a personalized quote