Is It Possible to Use Computer Methods

1
Is It Possible to Use Computer Methods to Get
the Information a Biopsy Provides without
Performing a Biopsy? Jim DeLeo, NIH Clinical
Center Facilitator
This or this?
Thursday, November 15, 2007 330 to 500
p.m. at the National Institutes of
Health Clinical Center (Building 10) Medical
Board Room (Room 2C116) This session is
sponsored by the NIH Biomedical Computing
Interest Group www.nih-bcig.org Additional
information 301-496-0191
2
The bad news is

• According to the Journal of the
• American Medical Association
• medical treatment is the third-
• leading cause of death.
• From A New Earth, by Eckhart Tolle

3
Unnecessary Biopsies

• Video Alert for unnecessary biopsies
• Revolutionary Breakthrough in Breast Cancer
  Detection... reduce the need for biopsies. In
  fact, the American Cancer Society confirms only
  20 percent of biopsies ...http//vids.myspace.com
  
• http//www.healthcentral.com/video/408/2051.html

4
Prostate Cancer

• Predictive computer model reduces unnecessary
  biopsies by more than one-third
• June 25, 2003. Using a predictive computer
  model could reduce
• unnecessary prostate biopsies by almost 38,
  according to a study
• conducted by Oregon Health Science University
  researchers. The
• study was presented at the American Society of
  Clinical Oncology's
• annual meeting in Chicago.
• "While current prostate cancer screening
  practices are good at
• helping us find patients with cancer, they
  unfortunately also identify
• many patients who don't have cancer. In fact,
  three out of four men who
• undergo a prostate biopsy do not have cancer at
  all," said Mark Garzotto,
• MD, lead study investigator and member of the
  OHSU Cancer Institute.
• "Until now most patients with abnormal screening
  results were counseled
• to have prostate biopsies because physicians were
  unable to
• discriminate between those with cancer and those
  without cancer."

5
Blood Tests Provide Alternative to Liver Biopsy
for Assessing Status of Liver Disease in HCV
Patients

• Oneida TheraDiagnostics Ltd has launched two
  non-invasive blood tests for assessing the extent
  of liver disease in patients infected with
  hepatitis C virus (HCV). Developed by
  hepatologists at the Pitie-Salpetriere Hospital
  and BioPredictive in France, the two tests are
  called FibroTest and ActiTest.
• The new tests provide easily accessible
  alternatives to liver biopsy, which is currently
  used to assess liver fibrosis and
  necroinflammatory activity in these patients. The
  tests do not replace the use of liver biopsy,
  however, which may be the appropriate diagnostic
  in many cases. Whether to use these new blood
  tests instead of a liver biopsy should be
  discussed between each individual patient and
  his/her primary care physician or liver
  specialist.
• Although liver biopsy has long been considered
  the gold standard for monitoring the status of
  HCV and to determine therapy options, it is an
  invasive procedure that carries some risk of
  serious complications. The new assays use a
  combination of six serum biochemical markers,
  plus age and gender data, in a patented algorithm
  to determine the degree of liver fibrosis and the
  level of ongoing necroinflammatory activity.

6
NIH Core Strategic Vision

• NIH Director Elias Zerhouni, M.D. has recently
  emphasized
• the following characteristics of the NIH Core
  Strategic
• Vision
• 1. to transform medicine and health from a
  curative to a preemptive paradigm.
• 2. to support basic research to identify the
  earliest molecular stages of disease in complex
  biological systems.
• 3. to accelerate translation of findings from
  the bench to the bedside to the community.
• 4. to provide the evidence and knowledge base to
  allow for a rational transformation of our
  healthcare system.

7
NIH Core Strategic Vision

• Predictive
• Preemptive Medicine
• Personalized

8

• Computer Science
• Data Mining
• Machine Learning
• Medical Composite Index
• Intelligent Virtual Biopsy

9
Data Mining
Data Preparation
Data Gathering
Data Preprocessing
Data Standard- ization
Data Selection
Data Production
Feature Extraction
Feature Normal- ization
Data Pooling
Knowledge Discovery
Basic Statistics
Data Visualization
Unsupervised Learning
Supervised Learning
Semi-supervised Learning
10
(No Transcript)
11
(No Transcript)
12
30 X2 5
n
d
d
n
n
d
d
n
d
n
n
d
d
n
n
d
n
d
n
d
n
100 500
X1
13
1 x2 0
30 X2 5
n
d
d
n
n
d
d
n
d
n
n
d
d
n
n
d
n
d
n
d
n
0
1 x1
100 500
X1
14
TRANSFORMATION
1 x2 0
30 X2 5
d
n
d
n
d
d
n
d
n
n
d
d
d
n
n
n
d
d
n
n
d
d
d
n
n
d
n
n
n
d
d
n
d
n
n
n
d
d
n
d
n
0
1 x1
100 500
X1
15
TRANSFORMATION
1 x2 0
30 X2 5
d
n
d
n
d
d
n
d
n
n
d
d
d
n
n
n
d
d
n
n
d
d
d
n
n
d
n
n
n
d
d
n
d
n
n
n
d
d
n
d
n
0
1 x1
100 500
X1
16
Other Considerations

• ROC Methodology
• Prevalence
• Misclassification Costs
• Ensembling
• Hybridized Methodologies

17
Simulated Data ROC Plot
18
PSA ROC Plots
Normal vs. Cancer Area 0.95
N/C
B/C
Benign vs. Cancer Area 0.79
sensitivity
1 - specificity
19
An Intelligent Agent
20
E N S E M B L E O U T P U T
P A T I E N T D A T A
21
Hybrid of Decision Trees and ANN Ensembles
SA1
SB2
SB1
DC1
DC2
SC1
SC2
DD1
DD2
DD3
DE6
DD4
DD5
22
Prostate Cancer Virtual Biopsy ANN
23
Biopsy Replacement Levels
24
Extreme Multidisciplinary Team Work
25
Scientific Computing Section Clinical Center,
National Institutes of Health, Bethesda,
MD Summer 2007
26
Idiopathic Inflammatory Myopathy Extreme
MultidisciplinaryTeam
Scientific Computing Section Clinical Center,
National Institutes of Health, Bethesda,
MD Summer 2007
27
Questions

• Is It Possible to Use Computer Methods
• to Get the Information a Biopsy Provides
• without Performing a Biopsy?
• What should we start with?
• - most likely to succeed application
• - where there exists data
• How do we proceed to rapidly move from
• bench to bedside to community with the
  first
• application?

28
Is It Possible to Use Computer Methods to Get
the Information a Biopsy Provides without
Performing a Biopsy? Jim DeLeo, NIH Clinical
Center Facilitator
This or this?
Thursday, November 15, 2007 330 to 500
p.m. at the National Institutes of
Health Clinical Center (Building 10) Medical
Board Room (Room 2C116) This session is
sponsored by the NIH Biomedical Computing
Interest Group www.nih-bcig.org Additional
information 301-496-0191