Docsity
Docsity

Prepare for your exams
Prepare for your exams

Study with the several resources on Docsity


Earn points to download
Earn points to download

Earn points by helping other students or get them with a premium plan


Guidelines and tips
Guidelines and tips

Toward the Automatic Assessment of Behavioral Distrubances ..., Exercises of Nursing

2Western Psychiatric Institute and Clinic, University of Pittsburgh, ... The psychosocial care of the nursing home resident is affected by inadequa-.

Typology: Exercises

2022/2023

Uploaded on 05/11/2023

kalia
kalia 🇺🇸

4

(7)

239 documents

1 / 5

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Toward the Automatic Assessment of Behavioral
Disturbances of Dementia
S. J. Allin1, A. Bharucha2, J. Zimmerman1, D. Wilson1, M. J. Roberson1, S.
Stevens1, H. Wactlar1and C.G. Atkeson1
1Carnegie Mellon University, Pittsburgh PA
2Western Psychiatric Institute and Clinic, University of Pittsburgh, Pittsburgh PA
1 An Example of Pervasive Computing in Healthcare
Advances in healthcare have dramatically increased life expectancy in the United
States over the course of the last century. The challenges of caring for a graying
population are exemplified by Alzheimer‘s disease (AD), the most common cause
of dementia among persons age 65 and older. The current annual direct and
indirect cost of 100 billion dollars devoted to dementia care will increase as the
number of Americans suffering from AD nearly triples to 14 million by the year
2050 [6].
Approximately 90% of individuals with dementia exhibit one or more be-
havioral disturbances over the course of the illness [3]. These behaviors can be
broadly classified as physically aggressive, physically non-aggressive (e.g., wan-
dering), verbally aggressive, and verbally non-aggressive (i.e., repetitive vocal-
izations). These symptoms not only reduce the quality of life of the individual
with dementia and his/her caregiver, but also lead to premature institutionaliza-
tion, inappropriate and excessive physical and chemical restraints and attendant
medical consequences such as falls.
The psychosocial care of the nursing home resident is affected by inadequa-
cies in staffing and training that are ubiquitous. As a result, clinical data often
consists of brief observations of residents over relatively short periods of time,
filtered through the lens of an overburdened staffmember. In the absence of
objective, reliable assessment and outcomes measurement methodologies, effec-
tiveness of behavioral and pharmacological interventions cannot be determined.
Pervasive technology holds the promise of developing objective, real-time,
continuous assessment and outcomes measurement methodologies that were pre-
viously unfeasible. Such technologies can contribute greatly to a deeper un-
derstanding of the activity and behavior patterns of individual residents, and
the physical, environmental and psychosocial correlates of these patterns. These
kinds of applications accord very well with research goals articulated by the
International Psychogeriatric Association in 2002 [1].
2 Current Work
In February of 2002, researchers at Carnegie Mellon University began a collab-
oration with Dr. Ash Bharucha, a geriatric psychiatrist from the University of
Fifth Int'l Conference on Ubiquitous Computing (UbiComp'03), 2nd Int'l Workshop on Ubiquitous Computing for Pervasive Healthcare Applications, Seattle, WA, Oct 12-15, 2003.
pf3
pf4
pf5

Partial preview of the text

Download Toward the Automatic Assessment of Behavioral Distrubances ... and more Exercises Nursing in PDF only on Docsity!

Toward the Automatic Assessment of Behavioral

Disturbances of Dementia

S. J. Allin^1 , A. Bharucha^2 , J. Zimmerman^1 , D. Wilson^1 , M. J. Roberson^1 , S. Stevens^1 , H. Wactlar^1 and C.G. Atkeson^1 (^1) Carnegie Mellon University, Pittsburgh PA (^2) Western Psychiatric Institute and Clinic, University of Pittsburgh, Pittsburgh PA

1 An Example of Pervasive Computing in Healthcare

Advances in healthcare have dramatically increased life expectancy in the United States over the course of the last century. The challenges of caring for a graying population are exemplified by Alzheimer‘s disease (AD), the most common cause of dementia among persons age 65 and older. The current annual direct and indirect cost of 100 billion dollars devoted to dementia care will increase as the number of Americans suffering from AD nearly triples to 14 million by the year 2050 [6]. Approximately 90% of individuals with dementia exhibit one or more be- havioral disturbances over the course of the illness [3]. These behaviors can be broadly classified as physically aggressive, physically non-aggressive (e.g., wan- dering), verbally aggressive, and verbally non-aggressive (i.e., repetitive vocal- izations). These symptoms not only reduce the quality of life of the individual with dementia and his/her caregiver, but also lead to premature institutionaliza- tion, inappropriate and excessive physical and chemical restraints and attendant medical consequences such as falls. The psychosocial care of the nursing home resident is affected by inadequa- cies in staffing and training that are ubiquitous. As a result, clinical data often consists of brief observations of residents over relatively short periods of time, filtered through the lens of an overburdened staff member. In the absence of objective, reliable assessment and outcomes measurement methodologies, effec- tiveness of behavioral and pharmacological interventions cannot be determined. Pervasive technology holds the promise of developing objective, real-time, continuous assessment and outcomes measurement methodologies that were pre- viously unfeasible. Such technologies can contribute greatly to a deeper un- derstanding of the activity and behavior patterns of individual residents, and the physical, environmental and psychosocial correlates of these patterns. These kinds of applications accord very well with research goals articulated by the International Psychogeriatric Association in 2002 [1].

2 Current Work

In February of 2002, researchers at Carnegie Mellon University began a collab- oration with Dr. Ash Bharucha, a geriatric psychiatrist from the University of

Pittsburgh‘s Western Psychiatric Institute and Clinic, to assess the utility of pervasive technology in a dementia unit for the elderly. The group initially in- strumented environments with cameras and microphones. This choice of sensors enables caregivers to see and hear recordings of episodes as directly as possible as they view statistical summaries. The group is also considering instrumenting wheelchairs, walkers, canes, and shoes to help identify, locate, and monitor pa- tients. Problems are expected with wearable sensors that are not hidden from the patients or that require their cooperation, however. Residents of a dementia unit sometimes repeatedly undress and will forget to wear a badge. In addition, sensors requiring significant alterations to an individual‘s environment or rou- tine may be less acceptable, especially among a population that is particularly sensitive to even small changes [4]. In the initial study, four cameras and microphones were mounted in the de- mentia unit of a local nursing home. The cameras recorded activity in public areas of the facility for eight hours a day over the course of one week. Four view- points were selected: one in the dining room, one in the main hallway adjacent to rooms, and the last two in the television room. Resulting videos were annotated by a group of undergraduate students under the supervision of Dr. Bharucha. Together, they scanned the data for visible or audible indicators of behavioral symptoms. These indicators were based upon those listed in the Cohen Mans- field Agitation Inventory as well as the Pittsburgh Agitation Scale. Both have established reliability and validity in the nursing home setting [1]. Based on the group‘s observation of videos, several initial goals for automa- tion were identified. These include the quantitative study of:

  1. Interpersonal interactions, such as when a patients talks with, is touched by or touches another.
  2. Changes in activity, such as when a typically active patient fails to appear in group activities, or at meals.
  3. Wandering, such as when a patient meanders into anothers room, appears lost, or makes an effort to escape.
  4. Physically and verbally aggressive actions, such as when a patient talks loudly, hits, kicks, or curses. Interactive designers led by John Zimmerman have been analyzing ways in which this kind of automation could ultimately serve caretakers. They conducted brainstorming sessions to identify discrete user groups and applications, and are following this up with an ethnographic study of caregivers in dementia wards. Prototype interfaces are now being constructed; these will be populated with data from the pilot study and evaluated by caregivers. An arm of research, then, has been aimed at discovering how proposed technologies can best meet the needs of caregivers. In addition, three tiers of technical research have been formulated to address articulated goals. These are:

As implemented, the tracker works sufficiently well to initialize a color target once motion has been detected, and to follow it. Yet errors still, inevitably, are made. The tracker occasionally comes to reside on objects in the background or becomes lost as people pass one another. Because of this, a part of our research seeks to determine what level of ‘ accuracy‘ is sufficient for our task. Can a tracker that falls off an individual 80% of the time yield reasonable estimates about his or her ability to ambulate? Can it be used to determine when he or she is trying to escape or in another persons room? Algorithms which operate on image and audio statistics at this level of gran- ularity may be best suited to detect instances of wandering and changes in activity. 2.3 Level Three: Use of Fine Grain Measurements and Statistics Pertaining to an Individual Much research devoted to recognizing an individual‘s activities makes use of models built atop on prior knowledge of typifying kinematics and dynamics [9]. A good deal more relies on models built atop prior knowledge of token actions and sequences of these tokens [2]. Such representations are powerful, yet the models are complicated, difficult to initialize automatically and unstable, especially when scenes are cluttered and reflect the activity of groups. At level three, researchers are seeking to make use of kinematic and proba- bilistic models to recognize detailed activities, actions, and events like falls. In addition, they seek to use models of affect and emotion that depend upon very high resolution images of faces, and high quality recordings of voices. In our next pilot study, plans have been made to utilize high resolution still cameras at din- ner tables and in doorways. The hope is to populate acquired data with images that may reveal subtle and semantically relevant attributes of faces and hands. Finally, color based segmentations are being used to segment heads from torsos, and torsos from arms and legs. Activities such as falls or hits may ultimately be encoded based on relationships between these various atomic parts; these rela- tionships can potentially be used to detect instances of physically aggressive actions, interpersonal interaction and changes in activity.

3 Expectations for the Workshop

The UBICOMP workshop provides the first author an opportunity to meet other researchers who are working with similar technologies and on similar is- sues. Moreover, it offers a valuable opportunity to achieve consensus among researchers as to problem areas in need of further research. How other researchers determine accuracy in pervasive systems and relate it to overarching system goals is of interest. It is also of interest to know how others take into account the privacy of patients and the possibility that systems may contribute to paranoia and insecurity. We recognize that our sensor choice represents one of many potential configurations, each with merits and drawbacks.

4 Research Activities, Including Biographies

Sonya Allin is an NSF Graduate Fellow pursuing a Ph.D. in the Human Com- puter Interaction Institute of CMU. Her research is in the application of image understanding to activity recognition among the elderly. She also helped to assess force feedback devices‘ potential to physically rehabilitate stroke survivors. Ashok Bharucha is a geriatric psychiatrist and Assistant Professor of Psychi- atry at the Western Psychiatric Institute and Clinic. He specializes in nursing home psychiatry with a research emphasis on the behavioral symptoms of de- mentia. John Zimmerman is an Assistant Professor at CMU with a joint appointment in the Human-Computer Interaction Institute and The School of Design. Daniel Wilson is a third year Ph.D. student in the Robotics Institute at CMU who researches the ability of simple sensors to predict identity and activity of home occupants. Matthew Johnson Roberson is in his third year of an undergraduate degree in Computer Science at CMU. Scott M. Stevens is Senior Systems Scientist in the School of Computer Sci- ence, CMU. He is a Co-PI of the Informedia Digital Library Project. Howard Wactlar is Vice Provost for Research Computing, Associate Dean, and Principal Research Scientist in the School of Computer Science, CMU. Chris Atkeson is a CMU Associate Professor in Robotics and HCI. His pre- vious projects include Georgia Tech’s Aware Home and Classroom 2000.

References

  1. Behavioral and psychological symptoms of demention educational pask. Interna- tional Psychogeriatric Association, 2002.
  2. M. Brand and I. Essa. Causal analysis for visual gesture understanding. AAAI Fall Symposium on Computational Models for Integrating Language and Vision,
  3. J. A. Brody. An epidemiologist views senile dementia: facts and figures. American Journal of Epidemiology, 113:155–162, 198 2.
  4. L. Burgio, K. Scilley, et al. Temporal patterns of disruptive vocalization in elderly nursing home residents. International Journal of Geriatric Psychiatry, 1 6:378-386,
  5. D. Commaniciu, V. Ramesh, and P. Meer. Real-time tracking of non-rigid objects using mean shift. CVPR, 2142-2145, 2000.
  6. J. W. Ernst, R. L. Hay and C. Fenn. Cognitive function and the costs of alzheimer’s disease: an exploratory study. Arch Neurol, 54(6):687-693, 1997.
  7. R. Jin and A. Hauptmann. Learning to identify video shots with people based on face detection. IEEE Int Conf on Multimedia and Expo, 2003.
  8. Y. Liu, R. Collins, and Y. Tsin. Gait sequence analysis using frieze patterns. ECCV, Copenhagen:657–671, 2002.
  9. H. Sidenbladh and M. Black. Learning the statistics of people in images and video. International Journal of Computer Vision, 54(1-3):183-2, 20 03.
  10. H. Zhang and J. Shi. Finding (un)usual events in video. Carnegie Mellon Univer- sity, CMU-RI-TR-03-0, 200 3.