Transcript slides - Georgetown University

Computer Science and Healthcare
Synergy
Howard Wactlar
III PI Meeting, April 2010
Carnegie Mellon University
Pittsburgh, USA
Lessons (being) Learned Collaborating
with Medical Practitioners
Copyright © 2010
Consider where in the research spectrum
• Quadrant model of scientific research (Stokes)
• One person’s basic research can be another person’s
application
Research
Inspiration
Consideration of use
NO
YES
YES
Pure Basic
Research
Bohr
Quest
For
Fundamental
Understanding
NO
Organizational
Data collection
Taxonomies
Use-inspired
Basic
Research
Understand
and control the
processes
Pasteur
Pure Applied
Research
Edison
Copyright © 2010
Different challenges from the same data
Automating the detection of behavioral & psychological symptoms of dementia
Computer Scientists:
Geriatric Psychiatrists:
• What are the health care applications of machine understanding of
video-based data?
• How well can we identify & track
individuals in real-world settings?
• How do we automate the recognition of activities, behaviors and
social interactions?
• How can we reduce and mine the
data so as to give healthcare
providers summaries of relevant
clinical events?
• How do we protect subjects’ privacy
and confidentiality?
• How do we develop continuous
capture and real-time processing
capabilities?
• How do we overcome?:
• Poor documentation
• Unreliable, uninformed
informants
• Biased reporting
• Cross-sectional observations
• How can we diagnose early and
accurately?
• How can we assess the safety
and efficacy of treatment
interventions?
• How can we assess the
implementation of those
recommendations?
• Evidence Based Medicine
Copyright © 2010
CareMedia: What are the observables?
• Who?
• Identify people across
cameras, days.
• What are they doing?
• Wandering around
• Socially interacting
• Looking for things
• Eating, sleeping in public
• How well did they do it?
• Quantify normal performance
/ measure change
• Detect/report anomalies
Click Here
Copyright © 2010
Labeling Complex Motions and Sequences
• Walking
• Approaching
• Standing
• Talking
• Hugging
• Hand touch body normally
• Shaking hands
• Walking (moving) together
• Hand in hand
Enable audio / Click Here
Copyright © 2010
Measure performance relevant to both disciplines
Automated recognition performance – for CS researchers
Training
Set
Test
Set
Recognition
Rate
False
Alarms
Passing
21
15
93%
4
Standing conversation
25
28
100%
7
Greeting
7
6
33%
2
Walking assistance
35
40
88%
4
Wheelchair pushing
5
4
75%
2
Encounter
59
65
94%
1
Interactions
Determine a domain to impact a documented problem
– for Medical researchers
Copyright © 2010
Operational Definition of Aggression
“An overt act, involving the delivery of noxious stimuli to
(but not necessarily aimed at) another object, organism
or self, which is clearly not accidental.”
Patel & Hope, Acta Psychiatr Scand 1992;85:131-135
AB = aggressive behavior
PAB = physically aggressive behavior
VAB = verbally aggressive behavior
Examples: spitting, grabbing, banging, pinching/squeezing,
punching, elbowing, slapping, tackling, using object as a weapon,
taking from others, kicking, scratching, throwing, knocking over,
pushing, pulling/tugging, biting, hurting self, obscene gesture, and
physically refusing care or activities
.
Copyright © 2010
Attempted Punch
Copyright © 2010
Hair Pulling
Copyright © 2010
Results of Aggression Recognition
• The top ten retrieval
results have an 80%
accuracy, which is much
better than the random
accuracy 36.2%
Copyright © 2010
The Healthcare Crisis
Copyright © 2010
The Good News
Copyright © 2010
The Good News
Copyright © 2010
The Good News
couple
Copyright © 2010
Population shift is coming, like it or not
!
Percent of US
population
70 and older:
UNITED STATES: 2000
9%
80+
75-79
70-74
65-69
60-64
55-59
50-54
45-49
40-44
35-39
30-34
25-29
20-24
15-19
10-14
5-9
0-4
MALE
14
12
10
8
6
4
2
0
FEMALE
0
2
Population (in millions)
Source: US Census Bureau, International database
16
Entire contents © 2006 Forrester Research, Inc. All rights reserved.
4
6
8
10
12
14
Population shift is coming, like it or not
Percent of US
population
70 and older:
!
UNITED STATES: 2050
16%
80+
75-79
70-74
65-69
60-64
55-59
50-54
45-49
40-44
35-39
30-34
25-29
20-24
15-19
10-14
5-9
0-4
MALE
14
12
10
8
6
4
2
0
20.4
FEMALE
0
2
Population (in millions)
Source: US Census Bureau, International database
17
Entire contents © 2006 Forrester Research, Inc. All rights reserved.
4
6
8
10
12
14
The Healthcare Crisis
• The most rapidly increasing age cohort is 85 and above.
• Nearly half of persons over age 85 have Alzheimer’s disease
• Disease prevalence with age > 85 years
•
•
•
•
Nursing home
Incontinence
Depression
Parkinson’s
20%
30%
10%
< 10%
• Comorbidity
• 80% have > 1 chronic condition
• 50% have > 2 chronic conditions
• 25% have > 3 chronic conditions
• For those >65, 30% of hospital admissions are due to
medication non-compliance
• By 2030, 1 in 2 working adults will be an informal caregiver
• This year the U.S. will graduate only 238 primary care
physicians
Copyright © 2010
The Healthcare Crisis (2)
• Its not just a cost crisis, it’s a capacity crisis
• The challenge for science and technology is to enable a
change in the healthcare delivery paradigm
• Home-centered healthcare: Move the care away from the hospital
/nursing home and the doctor / caregiver to the home and the individual
(+ partner) + technology
• This is not doing medicine. This is:
•
•
•
•
•
Sensing
Networking
Data mining
Predicting
Machine learning





Data collecting & securing
Information gathering & annotating
Correlating, summarizing & reporting
Behavior modification
Device actuating
Copyright © 2010
The Healthcare Crisis (3)
• Let’s restate this as a challenge:
• Move ¼ of institutional care to the home in 10 years
• Consider that as an appropriate III, HCC, and RI challenge
Copyright © 2010
Thank You
Questions ?
Howard Wactlar
III PI Meeting, April 2010
Carnegie Mellon University
Pittsburgh, USA