Ontology-based interactive visualization of patient-generated - - PowerPoint PPT Presentation

▶
ontology based interactive visualization of patient
SMART_READER_LITE
LIVE PREVIEW

Ontology-based interactive visualization of patient-generated - - PowerPoint PPT Presentation

Oct 27, 2023 269 likes •854 views

Ontology-based interactive visualization of patient-generated research questions David Borland RENCI, UNC-Chapel Hill Laura Christopherson RENCI, UNC-Chapel Hill Charles Schmitt NIEHS Overview Background Crohns disease &

slide-1
SLIDE 1

Ontology-based interactive visualization

  • f patient-generated research questions

David Borland Laura Christopherson Charles Schmitt RENCI, UNC-Chapel Hill RENCI, UNC-Chapel Hill NIEHS

slide-2
SLIDE 2

Overview

  • Background

– Crohn’s disease & colitis – IBD Partners discussion forum

  • Ontology creation
  • CCFA Explorer

– Ontology visualization – Research topic visualization

  • Overview
  • Details
  • Example scenario
  • Feedback and future work

Ontology-based interactive visualization of patient-generated research questions: Overview 2

slide-3
SLIDE 3

Crohn’s Disease and Colitis

  • Crohn’s disease

– Inflammatory bowel disease (IBD) – Chronic – Diverse set of symptoms

  • Diarrhea
  • Inflammation (gut and other body parts)
  • Fatigue
  • Abdominal pain
  • Weight loss
  • Etc.
  • Colitis

– Inflammation of the inner lining of the colon – Commonly co-occurs with Crohn’s disease

Ontology-based interactive visualization of patient-generated research questions: Background 3

slide-4
SLIDE 4

Crohn’s Disease and Colitis

  • Crohn’s disease

– Inflammatory bowel disease (IBD) – Chronic – Diverse set of symptoms

  • Diarrhea
  • Inflammation (gut and other body parts)
  • Fatigue
  • Abdominal pain
  • Weight loss
  • Etc.
  • Colitis

– Inflammation of the inner lining of the colon – Commonly co-occurs with Crohn’s disease

Ontology-based interactive visualization of patient-generated research questions: Background 4

No known cure Certain therapies can help treat symptoms Treatment largely consists

  • f disease management
slide-5
SLIDE 5

IBD Partners

(formerly CCFA Partners)

  • Patient-powered online research

network

– Crohn’s and Colitis Foundation – UNC-Chapel Hill School of Medicine

  • Discussion forum

– Patients propose and discuss research ideas – Researchers engage in discussion

Ontology-based interactive visualization of patient-generated research questions: Background 5

slide-6
SLIDE 6

Discussion Forum Data

Ontology-based interactive visualization of patient-generated research questions: Background 6

Research topic Research question Description User ID, Votes, Category Comment User ID Comment User ID

slide-7
SLIDE 7

Discussion Forum Data

Ontology-based interactive visualization of patient-generated research questions: Background 7

Research topic Research question Description User ID, Votes, Category Comment User ID Comment User ID Forum snapshot: 97 research topics 121 total comments 400 unique users 1246 total votes

slide-8
SLIDE 8

Discussion Forum Data

Ontology-based interactive visualization of patient-generated research questions: Background 8

Example topic post: Question:

Nicotine has shown to be effective for UC [ulcerative colitis] in some individuals, both prior- and nonsmokers. What is the mechanism? Does nicotine affect the microbiome, the immune system or both?

Description:

Big Pharma will not take on the role of studying nicotine as there is no $$$ in it. Few studies with small sample sizes have been done but more research is needed.

Research topic Research question Description User ID, Votes, Category Comment User ID Comment User ID

slide-9
SLIDE 9

Research topic Research question Description User ID, Votes, Category Comment User ID Comment User ID

Discussion Forum Data

Ontology-based interactive visualization of patient-generated research questions: Background 9

One of 9 predefined categories:

diet medications procedures and testing environment alternative therapies lifestyle genetics exercise

  • ther
slide-10
SLIDE 10

Research topic Research question Description User ID, Votes, Category Comment User ID Comment User ID

Discussion Forum Data

Ontology-based interactive visualization of patient-generated research questions: Background 10

Goals:

Identify common themes Prioritize patient-generated research questions

slide-11
SLIDE 11

Initial Approach

Ontology-based interactive visualization of patient-generated research questions: Background 11

Phrases were extracted from forum text (i.e., from questions, descriptions, and comments). Those phrases with >2 appearances were used to create the initial visualizations. Words Counts ulcerative colitis 20 in ammatory bowel 13 controlled trial 10 in ammatory bowel disease 10 bowel disease 10 top priority 10 periodontal disease 9 disease activity 8 vitamin d 7 ibd management 5

  • ther auto-immune

4 disease course 4 small bowel 4 combination therapy 4 This didn’t turn out to be as informative as we’d hoped. The frequency of words and/or word phrases didn’t successfully capture the ‘aboutness’ of the conversation on the CCFA forum.

slide-12
SLIDE 12

Initial Approach

Ontology-based interactive visualization of patient-generated research questions: Background 12

Phrases were extracted from forum text (i.e., from questions, descriptions, and comments). Those phrases with >2 appearances were used to create the initial visualizations. Words Counts ulcerative colitis 20 in ammatory bowel 13 controlled trial 10 in ammatory bowel disease 10 bowel disease 10 top priority 10 periodontal disease 9 disease activity 8 vitamin d 7 ibd management 5

  • ther auto-immune

4 disease course 4 small bowel 4 combination therapy 4 This didn’t turn out to be as informative as we’d hoped. The frequency of words and/or word phrases didn’t successfully capture the ‘aboutness’ of the conversation on the CCFA forum.

‘aboutness’

slide-13
SLIDE 13

Ontology Creation

  • Ontology

– Controlled vocabulary – Represent knowledge about a domain of interest – Enable multiple types of relationships

  • “is a”
  • “treats”
  • Etc.

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 13

slide-14
SLIDE 14

Content Analysis

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 14

slide-15
SLIDE 15

Content Analysis

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 15

The appearance of a given word in a text Conceptual and cannot be directly observed

Manifest Content vs. Latent Content

slide-16
SLIDE 16

Content Analysis

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 16

The appearance of a given word in a text Conceptual and cannot be directly observed E.g., “…the level of research anxiety present in user narratives about their experiences at the library.”

Wildemuth BM. Applications of Social Research Methods to Questions in Information and Library Science. Westport, Conn: Libraries Unlimited; 2009.

Manifest Content vs. Latent Content

slide-17
SLIDE 17

Content Analysis

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 17

Manifest Content vs. Latent Content

The appearance of a given word in a text Conceptual and cannot be directly observed E.g., “…the level of research anxiety present in user narratives about their experiences at the library.” “I am so anxious” “My heart won’t stop beating” “I wish I could relax”

Wildemuth BM. Applications of Social Research Methods to Questions in Information and Library Science. Westport, Conn: Libraries Unlimited; 2009.

slide-18
SLIDE 18

Content Analysis

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 18

1 2 4 8 8 10 14 16 34 5 10 15 20 25 30 35 40 Exercise Genetics Lifestyle Environment Alternative therapies Procedures and testing Medications Diet Other

Research Topic Count Category

slide-19
SLIDE 19

Content Analysis

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 19

Wildemuth BM. Applications of Social Research Methods to Questions in Information and Library Science. Westport, Conn: Libraries Unlimited; 2009.

“Sometimes there is no existing theory

  • r research on your message

populations; you may not know what the important variables are. The only way to discover them is to explore the content.”

1 2 4 8 8 10 14 16 34 5 10 15 20 25 30 35 40 Exercise Genetics Lifestyle Environment Alternative therapies Procedures and testing Medications Diet Other

Research Topic Count Category

slide-20
SLIDE 20

The Ontology

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 20

REPRESENTATIVE CLASSES POSTS comorbidity 11 diagnosis/monitoring method 7 IBD course 39 pre-diagnosis time period 1 diagnosis event 5 post-diagnosis time period 31 quality of life 8 risk factor 58 demographic factor 7 environmental factor 18 lifestyle factor 20 physiological factor 28 psychological factor 5 symptom 36 gastrointestional manifestation 12 extra-gastrointestinal manifestation 3 treatment method 50 alternative therapy 7 holistic treatment 12 medication 28 surgery 13

Protégé Borrows some terms from pre-existing ontologies:

Ontology of Adverse Events, Disease Ontology

slide-21
SLIDE 21

The Ontology

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 21

REPRESENTATIVE CLASSES POSTS comorbidity 11 diagnosis/monitoring method 7 IBD course 39 pre-diagnosis time period 1 diagnosis event 5 post-diagnosis time period 31 quality of life 8 risk factor 58 demographic factor 7 environmental factor 18 lifestyle factor 20 physiological factor 28 psychological factor 5 symptom 36 gastrointestional manifestation 12 extra-gastrointestinal manifestation 3 treatment method 50 alternative therapy 7 holistic treatment 12 medication 28 surgery 13

Protégé Borrows some terms from pre-existing ontologies:

Ontology of Adverse Events, Disease Ontology

337 classes 7 top-level classes

OWL→ ROBOT→ OBO Graph (JSON)

slide-22
SLIDE 22

The Ontology

Ontology-based interactive visualization of patient-generated research questions: Ontology creation 22

REPRESENTATIVE CLASSES POSTS comorbidity 11 diagnosis/monitoring method 7 IBD course 39 pre-diagnosis time period 1 diagnosis event 5 post-diagnosis time period 31 quality of life 8 risk factor 58 demographic factor 7 environmental factor 18 lifestyle factor 20 physiological factor 28 psychological factor 5 symptom 36 gastrointestional manifestation 12 extra-gastrointestinal manifestation 3 treatment method 50 alternative therapy 7 holistic treatment 12 medication 28 surgery 13

Protégé Borrows some terms from pre-existing ontologies:

Ontology of Adverse Events, Disease Ontology

337 classes 7 top-level classes

OWL→ ROBOT→ OBO Graph (JSON)

⟷ Research Topics Ontology Terms (Classes)

slide-23
SLIDE 23

CCFA Explorer

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 23

Ontology Visualization Topic Overview Topic Details

3 linked views D3 React/Flux Bootstrap

slide-24
SLIDE 24

Ontology Visualization

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 24

Topic Overview Topic Details

slide-25
SLIDE 25

Ontology Visualization

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 25

  • Force-directed network

– Node size

  • Number of topics labeled

with term

  • Labeled with child implies

labeled with parent

– Links

  • Class relationships
  • Currently only “is a”
slide-26
SLIDE 26

Topic Overview

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 26

Ontology Visualization Topic Details

slide-27
SLIDE 27

Topic Overview

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 27

  • t-SNE Layout

– Based on labeled ontology terms

  • Clusters of topics with similar
  • ntology terms
  • Glyph

– Size

  • Number of comments

– Border width

  • Number of votes

– Color

  • Category

van der Maaten LJP, Hinton GE. Visualizing High-Dimensional Data Using t-SNE. Journal of Machine Learning

  • Research. 2008;9:2579–2605.

diet environment alternative therapies medications procedures and testing lifestyle

  • ther

genetics exercise

slide-28
SLIDE 28

t-SNE Modifications

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 28

van der Maaten LJP, Hinton GE. Visualizing High-Dimensional Data Using t-SNE. Journal of Machine Learning

  • Research. 2008;9:2579–2605.
  • 1. Force-directed layout
  • f overlapping glyphs
slide-29
SLIDE 29

t-SNE Modifications

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 29

van der Maaten LJP, Hinton GE. Visualizing High-Dimensional Data Using t-SNE. Journal of Machine Learning

  • Research. 2008;9:2579–2605.
  • 2. Differential weighting of higher-level (left)
  • vs. lower level (right) ontology terms
slide-30
SLIDE 30

t-SNE Modifications

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 30

van der Maaten LJP, Hinton GE. Visualizing High-Dimensional Data Using t-SNE. Journal of Machine Learning

  • Research. 2008;9:2579–2605.
  • 3. Emphasizing selected
  • ntology terms for layout
slide-31
SLIDE 31

Topic Details

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 31

Ontology Visualization Topic Overview

slide-32
SLIDE 32

Topic Details

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 32

slide-33
SLIDE 33

Topic Details

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 33

slide-34
SLIDE 34

Topic Details

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 34

slide-35
SLIDE 35

Topic Details

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 35

  • List

– Sort

  • Topic ID
  • User ID
  • Votes
  • Comments
  • Category

– Filter

  • Selected topics
  • Selected ontology terms

– Highlight

  • Text from search box in red
slide-36
SLIDE 36

Interactive Selection and Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 36

  • Selection

– Ontology terms – Research topics – Selection in any view updates all views

  • Relationships

– Co-occurrence

  • Between ontology terms
  • Number of topics in common
  • Union for multiple

– Association

  • Between research topics
  • Number of terms in common
  • Union for multiple

– Connection

  • Between terms and topics
  • Topic labeled with term or not
  • Sum for multiple
slide-37
SLIDE 37

Interactive Selection and Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 37

  • Selection

– Ontology terms – Research topics – Selection in any view updates all views

  • Relationships

– Co-occurrence

  • Between ontology terms
  • Number of topics in common
  • Union for multiple

– Association

  • Between research topics
  • Number of terms in common
  • Union for multiple

– Connection

  • Between terms and topics
  • Topic labeled with term or not
  • Sum for multiple
slide-38
SLIDE 38

Interactive Selection and Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 38

  • Selection

– Ontology terms – Research topics – Selection in any view updates all views

  • Relationships

– Co-occurrence

  • Between ontology terms
  • Number of topics in common
  • Union for multiple

– Association

  • Between research topics
  • Number of terms in common
  • Union for multiple

– Connection

  • Between terms and topics
  • Topic labeled with term or not
  • Sum for multiple

Topic 1 Term 1, Term 2, Term 3 Topic 2 Term 1 Topic 3 Term 2, Term 3, Term 4

slide-39
SLIDE 39

Interactive Selection and Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 39

  • Selection

– Ontology terms – Research topics – Selection in any view updates all views

  • Relationships

– Co-occurrence

  • Between ontology terms
  • Number of topics in common
  • Union for multiple

– Association

  • Between research topics
  • Number of terms in common
  • Union for multiple

– Connection

  • Between terms and topics
  • Topic labeled with term or not
  • Sum for multiple

Topic 1 Term 1, Term 2, Term 3 Topic 2 Term 1 Topic 3 Term 2, Term 3, Term 4

Select Term 3 and Term 4

slide-40
SLIDE 40

Interactive Selection and Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 40

  • Selection

– Ontology terms – Research topics – Selection in any view updates all views

  • Relationships

– Co-occurrence

  • Between ontology terms
  • Number of topics in common
  • Union for multiple

– Association

  • Between research topics
  • Number of terms in common
  • Union for multiple

– Connection

  • Between terms and topics
  • Topic labeled with term or not
  • Sum for multiple

Topic 1 Term 1, Term 2, Term 3 Topic 2 Term 1 Topic 3 Term 2, Term 3, Term 4

Co-occurrence:

Term 1: 1 Term 2: 2 Term 3: 2 Term 4: 1

Select Term 3 and Term 4

slide-41
SLIDE 41

Interactive Selection and Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 41

  • Selection

– Ontology terms – Research topics – Selection in any view updates all views

  • Relationships

– Co-occurrence

  • Between ontology terms
  • Number of topics in common
  • Union for multiple

– Association

  • Between research topics
  • Number of terms in common
  • Union for multiple

– Connection

  • Between terms and topics
  • Topic labeled with term or not
  • Sum for multiple

Topic 1 Term 1, Term 2, Term 3 Topic 2 Term 1 Topic 3 Term 2, Term 3, Term 4

Co-occurrence:

Term 1: 1 Term 2: 2 Term 3: 2 Term 4: 1

Connection:

Topic 1: 1 Topic 2: 0 Topic 3: 2

Select Term 3 and Term 4

slide-42
SLIDE 42

Interactive Selection and Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 42

  • Selection

– Ontology terms – Research topics – Selection in any view updates all views

  • Relationships

– Co-occurrence

  • Between ontology terms
  • Number of topics in common
  • Union for multiple

– Association

  • Between research topics
  • Number of terms in common
  • Union for multiple

– Connection

  • Between terms and topics
  • Topic labeled with term or not
  • Sum for multiple

Topic 1 Term 1, Term 2, Term 3 Topic 2 Term 1 Topic 3 Term 2, Term 3, Term 4

Select Topic 2 and Topic 3

slide-43
SLIDE 43

Interactive Selection and Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 43

  • Selection

– Ontology terms – Research topics – Selection in any view updates all views

  • Relationships

– Co-occurrence

  • Between ontology terms
  • Number of topics in common
  • Union for multiple

– Association

  • Between research topics
  • Number of terms in common
  • Union for multiple

– Connection

  • Between terms and topics
  • Topic labeled with term or not
  • Sum for multiple

Topic 1 Term 1, Term 2, Term 3 Topic 2 Term 1 Topic 3 Term 2, Term 3, Term 4

Select Topic 2 and Topic 3 Association:

Topic 1: 3 Topic 2: 1 Topic 3: 3

slide-44
SLIDE 44

Interactive Selection and Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 44

  • Selection

– Ontology terms – Research topics – Selection in any view updates all views

  • Relationships

– Co-occurrence

  • Between ontology terms
  • Number of topics in common
  • Union for multiple

– Association

  • Between research topics
  • Number of terms in common
  • Union for multiple

– Connection

  • Between terms and topics
  • Topic labeled with term or not
  • Sum for multiple

Topic 1 Term 1, Term 2, Term 3 Topic 2 Term 1 Topic 3 Term 2, Term 3, Term 4

Select Topic 2 and Topic 3 Association:

Topic 1: 3 Topic 2: 1 Topic 3: 3

Connection:

Term 1: 1 Term 2: 1 Term 3: 1 Term 4: 1

slide-45
SLIDE 45

Ontology Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 45

Node interior size and color proportional to co-occurrence/connection Labels shown if > 25%

slide-46
SLIDE 46

Topic Overview Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 46

Glyph saturation proportional to association/connection

drug

slide-47
SLIDE 47

Topic Overview Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 47

Glyph saturation proportional to association/connection

drug drug + remission

slide-48
SLIDE 48

Topic Overview Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 48

Glyph saturation proportional to association/connection

drug drug + remission drug + remission weighted t-SNE

slide-49
SLIDE 49

Topic Details Highlighting

Ontology-based interactive visualization of patient-generated research questions: CCFA Explorer 49

Filtered by selected topics Ontology terms colored as in

  • ntology

network visualization

slide-50
SLIDE 50

Example Scenario

Ontology-based interactive visualization of patient-generated research questions: Example scenario 50

  • User searches for genetic makeup
  • Highlighted in red
  • User selects genetic makeup
  • Notices medication co-occurs with

genetic makeup

  • User select medication also
slide-51
SLIDE 51

Example Scenario

Ontology-based interactive visualization of patient-generated research questions: Example scenario 51

  • User re-runs t-SNE
  • Notices cluster of three topics,

including one very large glyph

  • User selects these three topics
slide-52
SLIDE 52

Example Scenario

Ontology-based interactive visualization of patient-generated research questions: Example scenario 52

  • User filters topic details to

show only these topics for in- depth inspection/comparison

slide-53
SLIDE 53

Feedback and Future Work

Ontology-based interactive visualization of patient-generated research questions: Feedback and future work 53

  • Demonstration to IBD Partners

– Generally positive feedback

  • Useful way to explore the forum data
  • Quickly identify major themes and popular research topics
  • Some features may be too complex for more naïve users

– Two particular themes identified

  • Patient-facing interface

– Find similar patients – More easily identify research topics relevant to them

  • Researcher-facing interface

– Quickly identify relevant information in their research area – Quickly generate summaries of relevant information – Easily presented to stakeholders

slide-54
SLIDE 54

Feedback and Future Work

Ontology-based interactive visualization of patient-generated research questions: Feedback and future work 54

  • Interface

– Redesign ontology visualization

  • Improve navigation

– Explore automatic text summarization

  • Include in summary panel for selected

terms/topics

  • Data processing

– Explore ways to automate/semi- automate topic classification – Enable ontology editing/expansion

  • User evaluations
slide-55
SLIDE 55

Feedback and Future Work

Ontology-based interactive visualization of patient-generated research questions: Feedback and future work 55

  • Interface

– Redesign ontology visualization

  • Improve navigation

– Explore automatic text summarization

  • Include in summary panel for selected

terms/topics

  • Data processing

– Explore ways to automate/semi- automate topic classification – Enable ontology editing/expansion

  • User evaluations

15,680 > 400

slide-56
SLIDE 56

Feedback and Future Work

Ontology-based interactive visualization of patient-generated research questions: Feedback and future work 56

  • Interface

– Redesign ontology visualization

  • Improve navigation

– Explore automatic text summarization

  • Include in summary panel for selected

terms/topics

  • Data processing

– Explore ways to automate/semi- automate topic classification – Enable ontology editing/expansion

  • User evaluations

15,680 > 400

slide-57
SLIDE 57

Questions?

Ontology-based interactive visualization of patient-generated research questions: Feedback and future work 57

Thanks to IBD Partners for their help with this work!