Amazon Comprehend Medical
A HIPAA-eligible NLP service for clinical text: medical entities, protected health information (PHI) detection, and linking to ICD-10-CM, RxNorm, and SNOMED CT, in real time or in batches, and how it differs from Amazon Comprehend and AWS HealthLake.
2 min read · Last reviewed · MLA-C02
Sign in to bookmark tutorials and track what you have read.
On this page
Already know this topic?
Try the Amazon Comprehend Medical quiz: 5 questions on Amazon Comprehend Medical.
What it is
Amazon Comprehend Medical uses natural language processing trained for healthcare to find information in unstructured clinical text: physician notes, discharge summaries, test results, and case notes. It detects medical entities (conditions, medications with dosage and frequency, tests, treatments, anatomy) and protected health information (PHI), and links entities to standard medical codes through ontology linking. It is a HIPAA (Health Insurance Portability and Accountability Act)A US law that protects health information. AWS customers can process protected health information on HIPAA-eligible services after accepting a Business Associate Addendum (BAA) with AWS, which is available in AWS Artifact.Glossary entry-eligible service, works on English (US) text, and doesn't persistently store your content.
Choose it when a question involves clinical or medical text: extracting diagnoses and medications, coding for billing, or finding PHI to redact before data is used for analytics or ML (Machine learning)Software that learns patterns from example data to make predictions or decisions, instead of following rules written by hand. A model is first trained on data, then used to make predictions on new data, which is called inference.Glossary entry (MLA-C02 tasks 1.2.8 and 2.1.8). For general text, use Amazon Comprehend.
Concepts and features
- Entity detection: medical conditions, medications, tests, treatments, procedures, anatomy, and their traits (for example negation: "no sign of pneumonia").
- PHI detection (
DetectPHI): names, dates, addresses, IDs, and other protected health information, to mask or remove before sharing data. - Ontology linking: InferICD10CM links conditions to ICD-10-CM codes, InferRxNorm links medications to RxNorm concepts, and InferSNOMEDCT links conditions, anatomy, tests, and procedures to SNOMED CT codes, each with confidence scores.
- Real-time API (Application programming interface)A defined way for one program to ask another for data or actions, such as a set of HTTPS requests and responses. Every AWS service is controlled through its API, which the console, the CLI, and the SDKs call for you.Glossary entry calls for single documents, and batch analysis jobs for documents in Amazon S3.
- Works with Amazon Transcribe (including Transcribe Medical) to analyze transcribed patient conversations, and with Amazon Textract for scanned documents.
Security, resilience, performance, and cost
Security
HIPAA eligible; all connections use TLS (Transport Layer Security)The protocol that encrypts traffic between a client and a server and lets the server prove its identity with a certificate. It replaced SSL, though the older name is still common; HTTPS is HTTP over TLS.Glossary entry; the service doesn't store your text, and batch input and output in S3 are encrypted with keys you control. Control access with IAM. Results need review by trained medical professionals before use in patient care.
Resilience
A managed API; batch jobs work through documents in S3 and write results there.
Performance
Use batch jobs for large document sets; set confidence thresholds per use case.
Cost
Pay per amount of text analyzed, by operation (entities, PHI, ontology linking), with no minimum fee; pricing differs from Amazon Comprehend.
Monitoring
CloudWatch metrics for API use; CloudTrail logs calls.
Comparisons
| Need | Choose |
|---|---|
| Entities, PHI, and medical codes from clinical notes | Amazon Comprehend Medical |
| Sentiment, entities, or PII (Personally identifiable information)Information that can identify a person, alone or combined with other data, such as a name, email address, or passport number. Amazon Macie finds PII in Amazon S3.Glossary entry in everyday text | Amazon Comprehend |
| Store and query health records in FHIR, with medical NLP built in | AWS HealthLake |
| Text from scanned forms | Amazon Textract, then Comprehend Medical |
Reverse mindmap: when the question points here
- Points to Comprehend Medical
- Diagnoses and medications from doctors' notes leads to Entity detection
- Remove PHI before training a model leads to DetectPHI
- Billing codes from clinical text leads to InferICD10CM
- Normalized drug names leads to InferRxNorm
- Points elsewhere
- Customer reviews' sentiment leads to Amazon Comprehend
- A FHIR data store for patient records leads to AWS HealthLake
Key takeaways
- Comprehend Medical extracts medical entities and PHI from English clinical text.
- Ontology linking maps to ICD-10-CM, RxNorm, and SNOMED CT.
- HIPAA eligible, no persistent storage of content; real-time and batch.
- Human review by medical professionals is required for patient care uses.