The situation
An experienced physician wanted to learn from past cases: which records looked like the patient in front of them, and what happened next.
Clinical decision support platform
A platform that finds similar medical records so doctors can reach clinical insights quickly and accurately.
A physician with 25 years of clinical experience
01 · Context
02 · Challenge
What had to be true
Pull the clinically meaningful details out of free-text records, remove protected health information, and match records on what matters.
03 · What we built
The system
- Entity extraction with AWS Comprehend Medical and SciSpacy NER: age, gender, diagnoses, conditions, symptoms, anatomy, procedures, and medications with dosage and strength
- Handling of clinical abbreviations, and removal of protected health information (PHI)
- Semantic and Elastic search to fetch similar medical records from these entities
- A follow-on project using OCR to extract patient, provider, service and payment details from prescriptions and invoices
04 Architecture
How it works, step by step.
Step through the system, or let it play.
- A medical record is loaded
- Clinical entities extracted
- PHI removed
- Semantic and Elastic search
- Similar records and insights
05 · Outcome
Doctors find similar records quickly and derive clinical insights with less manual searching.
06 · Stack
Next case · Education · Machine learningComputer adaptive testing for medical licensing exams