Paper: CareMedia: Automated Video and Sensor Analysis for Geriatric Care Bharucha, A.J., Under the panel, Innovations in Geriatric Care (Co-Chairs: Mark Miller, MD, Patricia Arean, PhD). Annual Meeting of the American Association for Geriatric Psychiatry, San Juan, Puerto Rico, March 10-13, 2006 Abstract:    Study: The design and implementation of specific non-pharmacological interventions in nursing homes has been hampered by the lack of sound understanding of the biopsychosocial and environmental context of behavioral disturbances. CareMedia is developing computer vision and machine learning technologies that capture in real-time, continuously an audiovisual and sensor record of activities, behaviors and social interactions of nursing home residents while simultaneously developing tools for automated data reduction and extraction, and safeguarding privacy. Methods: A feasibility study was conducted on a dementia unit by instrumenting the hallway and activity rooms with 4 surveillance caliber cameras and 8 audio-phones that stored data directly onto computer hard drives. The residents’ activities were recorded during four 2-hour blocks of time over ten days. Over 320 hours of video/audio data was processed with human coding of data and computer perception and machine learning techniques. Results: Eight (of 13) consenting residents were observed in the non-private spaces of the dementia unit for 13.6-24.6% of the time. Of this, interpersonal interactions of any kind (verbal or physical; with staff, visitors, other residents) were noted in < 20% of the recorded time. Meal times accounted for > 75% of the interpersonal interactions. A total of 7 bouts of physical aggression were observed, all accounted for by two residents, and preceded in a majority of cases by verbal aggression within 30 seconds of the events. Three of the 7 bouts of physical aggression were un-witnessed by staff, as were six elopements from the unit. Conclusion: Application of machine intelligence technologies as data collection, analysis and outcomes assessment tool in a nursing home environment is feasible, and will permit truly longitudinal, ecological studies of chronic care populations with continuous capture of objective measurements of their patterns of activities, behaviors and social interactions. Close