data-driven decision support to improve breast cancer
TRANSCRIPT
![Page 1: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/1.jpg)
Data-driven Decision Support
to Improve Breast Cancer
Diagnosis and Outcomes
Elizabeth Burnside, MD, MPH, MS Departments:
Radiology
Population Health
Biostatistics and Medical Informatics
Industrial and Systems Engineering
![Page 2: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/2.jpg)
Informatics & Breast Imaging
• General Overview
– History/Motivation
• Methodological Considerations
– Algorithms & metrics to measure performance
• Projects
– Improving mammographic predictions
– Improving image-guided core biopsy
![Page 3: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/3.jpg)
Motivation
• Information overload
– Medical articles in pubmed-online
– EHR information
– Genetic risk factors
• Human decision making involves
heuristics that may not scale up alone
![Page 4: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/4.jpg)
History
• Tools first conceived in:
– Leeds Abdominal Pain System went
operational in 1971 System = 91.8%
Physician = 79.6 % • Barriers
– Not integrated in clinical workflow
– Errors
![Page 5: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/5.jpg)
The Gail Model
• Uses data (BCDDP)
• Predicts Breast CA
– Five year/lifetime risk
Low signal
predictors
http://www.cancer.gov/bcrisktool/Default.aspx
![Page 6: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/6.jpg)
Predictive Information
Breast
Cancer
Age
![Page 7: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/7.jpg)
Human Computer Interaction
COMMUNICATION
Structured or Free Text
Report
Risk Score/
Probability
Risk Score/
Probability
![Page 8: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/8.jpg)
The Mammography Risk
Prediction Project
Elizabeth Burnside, MD, MPH, MS
C. David Page, PhD
Jude Shavlik, PhD
Charles Kahn, MD (MCW)
![Page 9: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/9.jpg)
Background-Opportunity
• 200,000 breast cancer diagnosed in US
• 20 million mammograms per year – False positives
• Millions of diagnostic mammograms/US
• Hundreds of thousands biopsies
– False negative • 10-30% of breast cancers not detected on mammography
• Variability of practice impacts many women
• Evidence-based decision support has the potential to drive substantial improvement
![Page 10: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/10.jpg)
Mass
Density -high
-equal
-low
-fat containing
Shape -round
-oval
-lobular
-irregular
Margins -circumscribed
-microlobulated
-obscured
-indistinct
-Spiculated Associated
Findings
Special
Cases
Architectural
Distortion Calcifications
Higher Probability
Malignancy
-fine pleomorphic
-linear/branching
Intermediate -amorphous
-course heterog
Typically Benign -skin
-vascular
-coarse/popcorn
-rod-like
-round
-lucent-centered
-eggshell/rim
-milk of calcium
-suture
-dystrophic
-punctate
BI-RADS
Trabecular
Thickening
Skin
Thickening
Nipple
Retraction
Skin
Retraction
Skin
Lesion
Axillary
Adenopathy
Focal Assymetric
Density
Assymetric
Breast Tissue
Lymph
Node
Tubular
Density
Distribution -clustered
-linear
-segmental
-regional
-diffuse/scattered
![Page 11: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/11.jpg)
Breast Cancer Probability Based
on BI-RADS Category
BI-RADS 0: Needs Additional Imaging
BI-RADS 1: Negative
BI-RADS 2: Benign
BI-RADS 3: Probably Benign
BI-RADS 4: Suspicious for malignancy
BI-RADS 5: Highly suggestive of malignancy
![Page 12: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/12.jpg)
Breast Cancer Predictors
BI-RADS
Category
![Page 13: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/13.jpg)
Breast Cancer Predictors
Detection Characterization Classification
Risk
![Page 14: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/14.jpg)
Bayesian Networks
Abnormal
mammo
Breast
Cancer
![Page 15: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/15.jpg)
What do we want to know?
We generally learn P(f|d)
Breast
cancer
We generally want to know P(d|f)
![Page 16: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/16.jpg)
Bayesian Networks
Abnormal
mammo
Breast
Cancer
Breast CA Symbol Prob
Yes P(d) 1.0%
No P(d-) 99.0%
Breast CA mammo Symbol Prob
Yes Yes P(f|d) 90%
Yes No P(f-|d) 10%
No Yes P(f|d-) 10.1%
No No P(f-|d-) 89.9%
P(f)
P(d) d)|P(f f)|P(d
![Page 17: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/17.jpg)
Estimates
Abnormal
mammo
Breast
Cancer
![Page 18: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/18.jpg)
Estimates
Abnormal
mammo
Breast
Cancer
8.3%
![Page 19: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/19.jpg)
Estimates
Abnormal
mammo
Breast
Cancer
![Page 20: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/20.jpg)
Estimates
Abnormal
mammo
Breast
Cancer
8.3%
![Page 21: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/21.jpg)
Probability Estimates
8.3% 8.3%
![Page 22: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/22.jpg)
Mass
Ca++ Lucent
Centered
Tubular
Density
Ca++ Amorphous
Ca++ Dystrophic
Mass Margins
Ca++ Round
Ca++ Punctate
Mass Size
Mass Stability Milk of
Calcium Ca++ Dermal
Ca++ Popcorn
Ca++ Fine/
Linear
Ca++ Eggshell
Ca++ Pleomorphic
Ca++ Rod-like
Skin Lesion
Architectural
Distortion
Mass Shape
Mass Density
Breast
Density
LN Asymmetric
Density
Age
HRT
FHx
Breast
Disease
![Page 23: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/23.jpg)
Pleomorphic microcalcifications
Clustered microcalcifications
Case
Example
Ductal Carcinoma in situ .48
Fibrocystic change .21
DC/DCIS .16
Ductal Carcinoma (NOS) .12
Malignant .760
Benign .239
Atypical .001
![Page 24: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/24.jpg)
Teaching cases
• 105 cases
• Created ROC curve
• Comparable to
– Neural networks
Az = .953
0
0.2
0.4
0.6
0.8
1
0 0.5 1
FPFT
PF
![Page 25: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/25.jpg)
Training on Data
• Motivation – Accurate probabilities are critical
– Some are not available in literature
– Modeling the relevant patient population is
possible with training
Expert &
Rule Based
Machine
Learning
![Page 26: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/26.jpg)
Mammogram Table
Mass shape
Mass margins
Mass density
Mass size
Mass stability
MicroCa++ shape
MicroCa++ distribution
BI-RADS category
…..
Patient Table
Age
Personal Hx Breast CA
Family Hx Breast CA
…..
Pathology Table
Pathology Result
Concordance
Recommendation
…..
Biopsy Table
Needle size
Number of samples
Post-proc appearance
Accurate clip position
…..
Registry Table
Patient ID
Margin status
Grade
Prior radiation
…..
Idea: Data Driven Decisions
![Page 27: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/27.jpg)
Patient Abnormality Date Calc Mass Mass
Size
location B/M
P1 1 5/08 N Y 3 mm RUO B
P1 2 5/10 Y Y 5 mm RUO M
P1 3 5/10 N Y 3 mm LLI B
P2 4 6/09 N Y N/A RLI B
… …
…
…
…
…
…
Idea: Data Driven Decisions
![Page 28: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/28.jpg)
Data
• Our dataset contains
–350 malignancies
–65,630 benign abnormalities
• Linked to cancer registry data
–Outcomes (benign/malignant)
![Page 29: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/29.jpg)
Training the BN
• Standard Machine learning
– Use known cases to train
– Use the tuning set for optimal training
– Performance based on hold out test set
Training
set
Tuning
set
Test
set
![Page 30: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/30.jpg)
• AUC 0.960 vs. 0.939 – P < 0.002
• Sensitivity – 90.0% vs. 85.3%
– P < 0.001
• Specificity – 93.9% vs. 88.1%
– P < 0.001
Performance
![Page 31: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/31.jpg)
What does that mean?
• At a specificity of 90%
38 conversions FN TP
• At a sensitivity of 85%
4226 conversions FP TN
![Page 32: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/32.jpg)
Ultimately Decision Support
Aids the Physician
• Output of the system is
– Advisory
– Utilized in the clinical context
– System performance alone is not the point
– Performance/Physician performance is the
key to improvement of care
![Page 33: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/33.jpg)
Collaborative Experiment
Radiologist
.916
Bayes Net
.919
Combined
.948
ROC curves
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1
FPF
TP
F
BN
Radiologist
Combined
![Page 34: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/34.jpg)
Results
Radiologist
.916
Bayes Net
.919
Combined
.948
ROC curves
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1
FPF
TP
F
BN
Radiologist
Combined
p=.03
![Page 35: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/35.jpg)
Results
Radiologist
.916
Bayes Net
.919
Combined
.948
ROC curves
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1
FPF
TP
F
BN
Radiologist
Combined p=.065
![Page 36: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/36.jpg)
Results
Radiologist
.916
Bayes Net
.919
Combined
.948
ROC curves
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1
FPF
TP
F
BN
Radiologist
Combined
p=.99
![Page 37: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/37.jpg)
UW Hospital
Same Testset / Different Training
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Recall
Pre
cis
ion
UW-Hospital Milwaukee Birads Both
Sensitivity
PP
V
Precision Recall Curves
![Page 38: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/38.jpg)
Calibration Curves
< 25%
25-50%
> 100%
50-75%
Observ
ed d
isease f
requency
Predicted Risk
< 25% 25-50% 50-75% >75%
![Page 39: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/39.jpg)
Calibration Curves
< 25%
25-50%
> 100%
50-75%
Observ
ed d
isease f
requency
Predicted Risk
< 25% 25-50% 50-75% >75%
![Page 40: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/40.jpg)
Calibration
Ayer, T., et al., Breast cancer risk estimation with artificial neural networks
revisited: discrimination and calibration. Cancer, 2010. 116(14): p. 3310-21.
• Hosemer-Lemishow
goodness of fit
![Page 41: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/41.jpg)
Mammogram Table
Mass shape
Mass margins
Mass density
Mass size
Mass stability
MicroCa++ shape
MicroCa++ distribution
BI-RADS category
…..
Patient Table
Age
Personal Hx Breast CA
Family Hx Breast CA
…..
Pathology Table
Pathology Result
Concordance
Recommendation
…..
Biopsy Table
Needle size
Number of samples
Post-proc appearance
Accurate clip position
…..
Registry Table
Patient ID
Margin status
Grade
Prior radiation
…..
Idea: Data Driven Decisions
![Page 42: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/42.jpg)
Patient Abnormality Date Calc Mass Mass
Size
location B/M
P1 1 5/08 N Y 3 mm RUO B
P1 2 5/10 Y Y 5 mm RUO M
P1 3 5/10 N Y 3 mm LLI B
P2 4 6/09 N Y N/A RLI B
… …
…
…
…
…
…
Idea: Data Driven Decisions
Not independent and identically distributed (IID)
![Page 43: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/43.jpg)
Algorithmic Opportunities
• Inductive logic programming (ILP)
• Statistical Relational Learning (SRL)
• Natural Language Processing (NLP)
![Page 44: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/44.jpg)
Algorithmic Opportunities
• Inductive logic programming (ILP)
• Statistical Relational Learning (SRL)
• Natural Language Processing (NLP)
![Page 45: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/45.jpg)
ILP
Abnormality A in
Mammogram M for Is malignant if:
Biopsy B in
Patient P
Malignant (A) IF
A has mass present
A has stability increasing
P has family history of breast cancer
B has atypia
![Page 46: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/46.jpg)
How does it work?
• Learn if-then rules that will become
features in a predictive model
– Inductive logic programming (ILP)
to learn the rules
– Integrated search strategy for constructing
and selecting rules for classifcation algorithm
![Page 47: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/47.jpg)
Human Computer Interaction
COMMUNICATION
Logical Rules
Logical Rules
![Page 48: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/48.jpg)
The Breast Biopsy Project
Elizabeth Burnside, MD, MPH, MS
Heather Neuman, MD, MS
Andreas Friedl, MD
C. David Page, PhD
Jude Shavlik, PhD
![Page 49: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/49.jpg)
Image-guided Breast Biopsy
Grobmyer, SR et al. Am J Surg. 2011 Feb 2.
Utilization of minimally invasive breast biopsy for
the evaluation of suspicious breast lesions.
• Excisional biopsy for
diagnosis of findings on
mammography is
overutilized
![Page 50: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/50.jpg)
Image-guided Breast Biopsy
• Core biopsy not perfect
– 10% of benign core biopsies
are non-definitive
– 10-15% of these are
upgraded to cancer at
excisional biopsy
Grobmyer, SR et al. Am J Surg. 2011 Feb 2.
Utilization of minimally invasive breast biopsy for
the evaluation of suspicious breast lesions.
![Page 51: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/51.jpg)
Breast Biopsy
• Biopsy: single most costly component of a
breast cancer screening program
• Annual breast biopsy utilization in 2010
62.6/10,000 women
700,000 women
~35,000-105,000 non-definitive
![Page 52: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/52.jpg)
Breast Biopsy at UW
• 5 year experience at UW
– 1228 consecutive image-guided core biopsies
• 890 benign
• 94 were deemed non-definitive
• 15 were upgraded to malignancy
• Hypothesis: ILP rules from the data and
from physicians could improve the
accuracy of upgrade prediction
![Page 53: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/53.jpg)
Biopsy data
• Example rule:
Upgrade (A) IF
concordance (A, d),
biopsyProcedure (A, US_core) and
pathDx (A, benign_breast_tissue)
• Incorporate physician and machine rules
into a Bayesian Network
![Page 54: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/54.jpg)
Physician rules
Machine rules
Evaluate
Incorporate
Evaluate
Incorporate
![Page 55: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/55.jpg)
6
5
BN
Machine generates
initial ILP rules
Data in
multiple
tables
Human generates
initial logic-based rules
MLN
Errors
Test on
new data
Rules incorporated
in graphical models
Machine & human
generate
new ILP rules
Repeated iterations to optimize performance
1b
2
3
7
Initial
logic-based
rule set
New
logic-based
rule set
1a
4
![Page 56: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/56.jpg)
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
0 0,2 0,4 0,6 0,8 1
Pre
cisi
on
(P
PV
)
Recall
Precision Recall Curve
WEKA-TAN
ILP rules alone
Human + ILP
![Page 57: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/57.jpg)
PPV Improvement
Baseline BN with rules
Benign biopsy 890 890
Non-definitive biopsy 94 75
Excision avoided 0 19
Malignant excision 15 15*
Benign excision 79 60
PPV 16.0% 20.0%
*No cancers missed
![Page 58: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/58.jpg)
Potential for Translation
• Translate these decision support
algorithms to the clinic to improve care
• Improve evidence-based decisions
• Encourage shared decision-making
![Page 59: Data-driven Decision Support to Improve Breast Cancer](https://reader031.vdocuments.us/reader031/viewer/2022020707/61fe086510b7ec6415129a06/html5/thumbnails/59.jpg)
Questions?