Summarizing medical documents using artificial intelligence and machine learning

Partner Overview

PareIT uses proprietary artificial intelligence and machine learning algorithms to sort legal case files and extract relevant information from medical records, including, diagnoses, prognoses, treatments, and medical history. PareIT has been trained on millions of records, ensuring both accuracy and efficiency in analysis. As a result, PareIT can analyze a thousand pages of medical records in less than a couple of hours. 

Challenge 

The aim was to develop a system for analyzing and summarizing medical documents with an AI/ML engine for end-users: lawyers, law firms, and insurance companies. The goal of the system was to provide a better, automated solution to the alternative provided by competitors of using low-wage workers for manual summaries.

Solution

The deliverables for this project were modules of the core analysis system that either enhance medical and personal information extraction, or solve a specific problem related to a domain: 

• Document Splitter – a system for splitting PDFs that have more than one document merged in them, using AWS Sagemaker
• Cataloguer – a module for document categorization and filtering non-medical documents, using AWS Sagemaker 
• Smart Section Search – a module for document structure and semantics analysis that separates documents into sections by structure and topic
• Entity Extraction – a module for personal and medical information retrieval
• Questionnaire analysis module – a module for analyzing specific page types 
• Encounter detection module – a module for separating information extracted from the document into sections based on the patient encounter
• BI using QuickSight 

The landing zone establishes a secure, scalable foundation for AWS. The landing zone includes AWS Control Tower, AWS Security Hub, AWS CloudTrail, AWS Config, AWS GuardDuty, AWS VPC, AWS Service Catalog, and other AWS services 

The MLOps pipelines ensure the easy building, training, and deployment processes for AI/ML models. The solution includes AWS SageMaker Pipelines, AWS SageMaker Training, and Inferences, AWSCodePipeline, AWS CodeBuild, and more. 

The self-service portal provisions infrastructure patterns with AWS Service Catalog. The portfolio includes MLOps pipelines, VPC products, and AWS Account Factory.

AWS Organisation and Cloud Foundation

VPC

MLOps Pipeline

Document Analysis ML Pipeline

Data Annotation

Application Architecture

Results 

• The solution provides faster and cheaper summarization than other competitor solutions using low-wage workers for manual summaries
• The team increased deployment frequency, reduced lead times, and decreased failure rate 
• The environment meets security and compliance requirements

About Neurons Lab

Neurons Lab is an AI consultancy that provides end-to-end services for mid and large-sized enterprises – from identifying high-impact AI applications to integrating and scaling the technology. We empower companies to capitalize on AI’s capabilities.

As an AWS Advanced Tier Services Partner and AWS Generative AI competency holder, our worldwide team includes data scientists, subject matter experts, and cloud specialists, complemented by an extensive talent pool of 500 experts. This expertise empowers us to tackle the most complex AI challenges and deliver solutions with exceptional speed, addressing both immediate needs and long-term strategic goals.

Ready to leverage AI for your business? Get in touch with the team here.