preha Establishing Precision Rehabilitation with Visual Analytics - - PowerPoint PPT Presentation

▶
preha
SMART_READER_LITE
LIVE PREVIEW

preha Establishing Precision Rehabilitation with Visual Analytics - - PowerPoint PPT Presentation

Nov 10, 2022 1.76k likes •2.15k views

preha Establishing Precision Rehabilitation with Visual Analytics Georg Bernold, Kresimir Matkovic, M.Eduard Grller, Renata G. Raidou Renata Raidou 2 Conventional Rehabilitation Renata Raidou 3 Precision Rehabilitation Challenges: Data

slide-1
SLIDE 1

preha

Establishing Precision Rehabilitation with Visual Analytics

Georg Bernold, Kresimir Matkovic, M.Eduard Gröller, Renata G. Raidou

slide-2
SLIDE 2

2 Renata Raidou

slide-3
SLIDE 3

Conventional Rehabilitation

3 Renata Raidou

slide-4
SLIDE 4

Precision Rehabilitation

Challenges: Data Resources Users Tasks

Renata Raidou 4

slide-5
SLIDE 5

Contribution

preha: a new approach to tackle the analysis of precision rehabilitation data.

Two main components:

1.

A detailed data–users–tasks analysis

2.

A visual analytics dashboard approach within preha

5 Renata Raidou

slide-6
SLIDE 6

Data–Users–Tasks Analysis

6 Renata Raidou

slide-7
SLIDE 7

Data–Users–Tasks Analysis

7

large – heterogeneous – high-dimensional1 – inconsistent2 – missing3 46,000 cases 2012 – 2019

Renata Raidou

slide-8
SLIDE 8

Data–Users–Tasks Analysis

8

Data Analysts Domain Experts Engineers

Renata Raidou

slide-9
SLIDE 9

Data–Users–Tasks Analysis

9

Interviews Abstract Tasks 30-50 minutes semi-structured

Renata Raidou

typologies for each task

slide-10
SLIDE 10

Data–Users–Tasks Analysis

Eng1: Provide meaningful data partitions Eng2: Prepare templates for patient assessment Eng3: Prepare templates for clinical benchmarking Eng4: Predict rehabilitation outcome Exp1: Show rehabilitation outcome to patients Exp2: Perform clinical benchmarking Exp3: Explore clinical datasets Exp4: Analyze data for clinical studies Exp5: Intervention planning

10 Renata Raidou

slide-11
SLIDE 11

Data–Users–Tasks Analysis

Eng1: Provide meaningful data partitions Eng2: Prepare templates for patient assessment Eng3: Prepare templates for clinical benchmarking Eng4: Predict rehabilitation outcome Exp1: Show rehabilitation outcome to patients Exp2: Perform clinical benchmarking Exp3: Explore clinical datasets Exp4: Analyze data for clinical studies Exp5: Intervention planning

11 Renata Raidou

slide-12
SLIDE 12

Renata Raidou

Eng4: Predict Rehabilitation Outcome

12

What? Why? How?

[inspired by Brehmer et al. 2013]

slide-13
SLIDE 13

Typologies for All Tasks

Renata Raidou 13

slide-14
SLIDE 14

preha

14 Renata Raidou

* if required by task

slide-15
SLIDE 15

preha

15 Renata Raidou

slide-16
SLIDE 16

Preprocessing

Rule-based approach, done once: easy to introduce new rules Profiling: identification and communication of quality problems

Set of regular expressions/rules defined by the users Whatever doesn’t match these “dirty”

Wrangling: modifying structure to make it suitable for processing

Standardization of tables and scores Each patient is assigned one (non-redundant) row in a data table

Cleansing: correcting dirty data

We know how correct data should look like Cleansing programs/rules to match this appearance

16 Renata Raidou

slide-17
SLIDE 17

preha

17 Renata Raidou

slide-18
SLIDE 18

preha

18 Renata Raidou

Visualization

slide-19
SLIDE 19

Visualization

Flexible, reusable, adaptable, expressive

Kibana framework:

All basic visualizations Extensible through d3.js Supports multiple linked views Interaction functionality Predictive analysis support

19 Renata Raidou

slide-20
SLIDE 20

Eng4: Predict Rehabilitation Outcome

20

What? Why? How?

Renata Raidou

[inspired by Brehmer et al. 2013]

slide-21
SLIDE 21

Predict assessment scores Visualize interesting characteristics of cohort Visualize interesting characteristics of cohort Filter Visualize interesting characteristics of cohort

slide-22
SLIDE 22

Predict assessment scores Visualize interesting characteristics of cohort Visualize interesting characteristics of cohort Filter Visualize interesting characteristics of cohort

slide-23
SLIDE 23

Visualize interesting characteristics of cohort Visualize interesting characteristics of cohort Filter Visualize interesting characteristics of cohort

slide-24
SLIDE 24

Visualize interesting characteristics of cohort Visualize interesting characteristics of cohort

Filter

Visualize interesting characteristics of cohort

slide-25
SLIDE 25

Filter

slide-26
SLIDE 26

Filter

slide-27
SLIDE 27
slide-28
SLIDE 28

Eng4: Use machine learning to predict rehabilitation outcome

slide-29
SLIDE 29

Dashboards for All Tasks

Renata Raidou 29

slide-30
SLIDE 30

Pilot Study

30

Introduce preha to four potential users Provide a set of small assignments to complete Findings:

Renata Raidou

Flexibility, adaptability to own working style Documentation/language, more digestible for domain experts (+) (-)

slide-31
SLIDE 31

Conclusion and Future Work

31

Design study for the workflow of precision rehabilitation Development of a dashboard-based strategy Extend evaluation to domain experts Predictive analytics extension Guided analytics incorporation

Renata Raidou

slide-32
SLIDE 32

32 rraidou@cg.tuwien.ac.at

Thank You! Questions?

preha

Georg Bernold, Kresimir Matkovic, M.Eduard Gröller, Renata G. Raidou