Medical records processing · Production

Advent Health Partners uses active learning and automation to speed up medical record labeling with Labelbox

The problem

Extracting data from millions of paper medical records to train ML models was slow and expensive because records regularly exceed 500 or even thousands of pages and arrive from numerous providers, hospitals, and insurers with no consistent formatting in mixed formats such as PDFs, faxes, and scans.

Workflow diagram · grounded in source
1
Medical records ingestion
trigger
“these records typically come from numerous different providers, hospitals, insurance companies, with no consistent formatting and contain image and text files of varying formats such as PDFs, faxes, scans, etc.”
2
OCR and NLP entity extraction
ai_action
“leveraging the latest advances in optical character recognition (OCR) on these medical records and applying AI through the use of natural language processing (NLP) in the form of entity extraction”
3
Automated page classification
ai_action
“they took paper records and automatically tagged emergency department records, discharge notes, and types of documents that human reviewers can check within the Labelbox interface”
4
Human review of model predictions
human_review
“Their in-house team of domain experts would review both confident model predictions to prevent errors and utilize uncertainty sampling”
5
Entropy-based active learning sampling
ai_action
“evaluated their existing model over unlabeled samples and calculated the entropy of the output classification vector. The AHP team grouped these calculations into two buckets: one with low entropy (meaning where the model was confident o…”
6
Semi-supervised class balancing
ai_action
“they used the earlier version of their model to create a semi-supervised model that would balance the classes within their unlabeled data which resulted in better model performance”
7
Labeled data fed to clinical ML models
output
“The AHP data science team's primary goal is to analyze and feed this vast amount of information into their clinical machine learning models for focused review”
Reported outcome

By leveraging Labelbox's active learning and model-assisted labeling, AHP dramatically sped up medical record labeling, reducing average time per label from 13 seconds to 8 seconds and cutting 25 hours off each specific labeling task.

Reported metrics
Average time per label — before13 seconds
Average time per label — after8 seconds
Labeling task time reduction per task25 hours
Labeling process speeddramatically speed up the labeling process
Reported stack
LabelboxOCRNLPmodel-assisted labeling
Source
https://labelbox.com/customers/advent-health-partners-customer-story/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

By leveraging Labelbox's active learning and model-assisted labeling, AHP dramatically sped up medical record labeling, reducing average time per label from 13 seconds to 8 seconds and cutting 25 hours off each specif…

What tools did this team use?

Labelbox, OCR, NLP, model-assisted labeling.

What results were reported?

Average time per label — before: 13 seconds; Average time per label — after: 8 seconds; Labeling task time reduction per task: 25 hours; Labeling process speed: dramatically speed up the labeling process (source-reported, not independently verified).

How is this medical records processing AI workflow structured?

Medical records ingestion → OCR and NLP entity extraction → Automated page classification → Human review of model predictions → Entropy-based active learning sampling → Semi-supervised class balancing → Labeled data fed to clinical ML models.

Related medical records processing cases