#01 Natural Language Processing for Healthcare - Named Entity Recognition

In this episode we explore named entity recognition (NER) and its uses in clustering 1 million hospital inpatients, monitoring pandemics and outbreaks, automating clinical coding, enriching research cohorts, and more.

Dev&Doc is a podcast where doctors and developers deep dive into the potential of AI in healthcare.
👨🏻‍⚕️Doc - Dr. Joshua Au Yeung
🤖Dev - Zeljko Kraljevic
LinkedIn Newsletter
YouTube
Spotify
Apple
Substack
For enquiries - 📧 Devanddoc@gmail.com

Timestamps:
00:00 Start
00:38 Intro
01:03 Setting the scene, clinical audit
03:13 What is Named Entity Recognition (NER)
14:59 Medical text as its own language
16:43 Medical abbreviations test
19:23 NER in different industries
21:55 NER with neural networks, deep learning, large language models
24:25 MedCAT medical concept annotation tool
25:50 When AI models go wrong, women get erectile dysfunction
28:20 Teaching a model to disambiguate
31:12 NER use case 1 - Clinical audit
33:04 How to fine tune a clinical model with clinician knowledge
36:07 NER use case 2 - Automating clinical audits
37:13 Why is NER not being used in the NHS? Windows XP
40:19 NHS is resistant to change
42:15 NER use case 3 - Enriching research databases
44:57 Which model should I use?
47:02 NER use case 4 - Extracting diseases from 1 million patients in King's College Hospital
52:05 Clustering 1 million patients with AI
55:14 Top 10 diagnoses in South London
58:45 Diseases by age in MIMIC dataset
1:01:27 Monitoring pandemic outbreaks
1:04:30 Predicting the future with Foresight

References:
Using machine learning for automated auditing of stroke comorbidities
Hospital-wide natural language processing summarising the health data of 1 million patients

🎞️ Editor - Dragan Kraljević
🎨 Brand design and art direction - Ana Grigorovici