AI for healthcare, AI for drug development, clinical trials
COMPOSE paper:
Doctor2Vec paper:
Table of Contents:
00:00 – Recent Innovation in Deep Learning for Clinical Trials
00:38 – Traditional Drug Discovery & Development Process
01:13 –
02:28 –
02:52 – Doctor Representation Learning for Doctor Selection
05:24 – Doctor2Vec: Overview
05:54 – Hierarchical Patient Embedding and Multimodal Trial Embedding
07:15 – Doctor2Vec: Dynamic Doctor Memory Network
09:34 – Data and Metrics
10:39 – Baseline
11:34 – Results
12:41 –
13:31 –
15:16 –
15:46 –
16:26 –
16:55 –
17:40 –
17:56 –
18:24 –
18:54 –
20:12 – The Patient-Trial Matching Task
22:32 – Challenges of The Patient-Trial Matching Task
25:02 – Method Overview: COMPOSE
26:01 – Method: Trial eligibility criteria embedding
27:36 – Method: Taxonomy Guided Patient Embedding
28:45 – Method: Trial eligibility criteria embedding
28:46 – Method: Taxonomy Guided Patient Embedding
28:53 – Method: Taxonomy Guided Patient Embedding
29:52 – Method: Attentional Record Alignment and Dynamic Matching
31:00 – Method: Explicit Inclusion/Exclusion Criteria Handling
32:35 – Performance Results
33:30 – Case Study: Attention Weights on Memory Slots
35:09 – Case Study: Failed Case
36:16 –
38:26 –
source
