Dipole : Diagnosis Prediction in Healthcare via Attention- based - - PowerPoint PPT Presentation

▶
dipole diagnosis prediction in healthcare via attention
SMART_READER_LITE
LIVE PREVIEW

Dipole : Diagnosis Prediction in Healthcare via Attention- based - - PowerPoint PPT Presentation

Nov 19, 2022 18 likes •250 views

Dipole : Diagnosis Prediction in Healthcare via Attention- based Bidirectional Recurrent Neural Networks Author: Fenglong Ma, Radha Chitta, Jing Zhou, Quanzeng You, Tong Sun, Jing Gao Source: KDD '17 Advisor: Jia-Ling Koh Speaker: Shih-Han Lo

slide-1
SLIDE 1

Dipole: Diagnosis Prediction in Healthcare via Attention- based Bidirectional Recurrent Neural Networks

Author: Fenglong Ma, Radha Chitta, Jing Zhou, Quanzeng You, Tong Sun, Jing Gao

Source: KDD '17 Advisor: Jia-Ling Koh Speaker: Shih-Han Lo Date: 2018/06/05

1

slide-2
SLIDE 2

Outline

  • Introduction
  • Method
  • Experiment
  • Conclusion

2

slide-3
SLIDE 3

Motivation

3

slide-4
SLIDE 4

Goal

4

Electronic Health Records Learn interpretable representations

Improve the accuracy, provide better interpretation (Especially when the length of the visits is large)

slide-5
SLIDE 5

Basic Notations

  • All unique medical codes from the EHR data:
  • Sequence of visits from the n-th patient:
  • Binary vector of medical codes:
  • Category representation:

5

slide-6
SLIDE 6

Outline

  • Introduction
  • Method
  • Experiment
  • Conclusion

6

slide-7
SLIDE 7

Proposed Model

7

slide-8
SLIDE 8

Visit Embedding

  • Given a visit 𝒚𝑗, we can obtain its vector representation as

follows:

8

slide-9
SLIDE 9

Bidirectional Recurrent Neural Networks

9

  • Structure of unidirectional RNN:
slide-10
SLIDE 10
  • Structure of bidirectional RNN:

Bidirectional Recurrent Neural Networks

10 𝑔 𝑔

slide-11
SLIDE 11

Attention Mechanism

  • 1. Location-based Attention:
  • 2. General Attention:
  • 3. Concatenation-based Attention:

11

slide-12
SLIDE 12

Attention Mechanism

  • Attention weight vector 𝜷𝑢:
  • Context vector 𝒅𝑢:

12

slide-13
SLIDE 13

Diagnosis Prediction

  • Attentional hidden state (attentional vector):
  • The attentional vector is fed to the softmax layer to produce

(t+1)-th visit information:

13

slide-14
SLIDE 14

Objective Function

  • Cross entropy to calculate the loss for all the patients:

14

Ground truth visit Predicted visit

slide-15
SLIDE 15

Interpretation

  • The top k codes with the largest values are selected:

15

slide-16
SLIDE 16

Interpretation

16

slide-17
SLIDE 17

Outline

  • Introduction
  • Method
  • Experiment
  • Conclusion

17

slide-18
SLIDE 18

Datasets

18

slide-19
SLIDE 19

Results of Diagnosis Prediction

19

slide-20
SLIDE 20

Results of Diagnosis Prediction

20

slide-21
SLIDE 21

Assumption Validation

21

slide-22
SLIDE 22

Outline

  • Introduction
  • Method
  • Experiment
  • Conclusion

22

slide-23
SLIDE 23

Conclusion

  • By employing bidirectional recurrent neural networks (BRNN),

Dipole can remember the hidden knowledge learned from the previous and future visits.

  • Attention Mechanisms allow us to interpret the prediction results

reasonably.

23