Computer Games for Older Adults Beyond Entertainment and Training: Possible Tools
for Early Warnings
Budapest University of Technology and Economics
04 Dec, 2015
2/19
Motivation Possible approaches Conceptual model Challenges How to assess the cognitive state Proof of the concept ICT architecture and components Conclusion and future work
Outline
22 May, 2015 ICT4AgeingWell Conference, 20-22 May, 2015
3/19
Primary goal: early detection of Mild Cognitive Impairment (MCI) (+ training) – WHY? Dementia is one of the main causes of dependency MCI: conversion rate to dementia is much higher Detection and treatment in the early phase can
delay the progression Hard to identify when the natural decline becomes
abnormal MCI is not a specific disease – problems with
memory,attention,visuospatial skills (the ability to interpret
objects and shapes)
Motivation (M3W project AAL Joint Progr.)
22 May, 2015 ICT4AgeingWell Conference, 20-22 May, 2015
planninglanguage
4/19
Cognitive tests: only when there are clear signs of cognitive deficit early detection is rare
Traditional, validated, paper-based clinical tests require specialist centres and highly trained professionals
Growing interest in the development of special computer games for cognitive monitoring and training purposes
Motivation
22 May, 2015 ICT4AgeingWell Conference, 20-22 May, 2015
5/19
Game development policy: Adapting well-known, popular games Transforming special clinical tests (Paired
Associates Learning, Corsi block-tapping) Developing new games specially designed
for this purpose
Monitoring policy: Voluntary Controlled
Possible approaches(our approach)
22 May, 2015 ICT4AgeingWell Conference, 20-22 May, 2015
6/19
Assessment of the mental state On an absolute scale Only significant change is to be detected
Reference for change detection Comparison to a reference group (inter-
personal) Comparison to a previously measured
reference performance of the same person
22 May, 2015
Possible approaches
7/19
With regular but voluntary use of computer games developed or modified specifically for older adults, we may be able to measure the mental changes and tendencies over time in an entertaining way.
Conceptual model
22 May, 2015 ICT4AgeingWell Conference, 20-22 May, 2015
8/19
Typical performance series of a player measured with a given computer game
22 May, 2015
Conceptual model
ICT4AgeingWell Conference, 20-22 May, 2015
9/19
How to measure cognitive performance using computer games?
How to cope with the typically heavy noise caused by the uncontrolled (home) measurement environment?
How to motivate people to take part in the long run?
How to compare performance shown in different games, which is basically a special sensor-fusion problem?
Challenges
22 May, 2015 ICT4AgeingWell Conference, 20-22 May, 2015
10/19
entertainment capability measurement power (contradictory requirements )
Challenges
22 May, 2015
GAMENUM. OF GAMELOGS
OPENED (June 01, 2014-March 31, 2015)
GIVEN UP
GIVEN UP %
NUM. OF GAMELOGS (GIVEN
UP NOT INCL.) (June 01, 2014-March 31, 2015)
NUM. OF SUBTYPES
PLAYED
NUM. OF CORRECT LOGS
(NO GIVEN UP)MOST FREQUENT
SUBTYPE
NUM. OF PLAYERS(MOST
FREQ.SUBTYPE)
STDPlaytime / AVGPlayTime
(BASED ON TOP 10 PLAYERS)
PLANAR 17 495 1 535 8.8 15 960 10 10 736 225 0.84
MEMORY 16 408 1 700 10.4 14 708 22 7 864 257 0.30
WGUESS 14 194 629 4.4 13 564 65 8 167 162 0.57
BLOCKS 10 415 830 8.0 9 585 5 6 358 202 0.49
CONNECT 10 276 1 519 14.8 8 757 12 6 079 167 0.65HIDDEN 14 563 6 002 41.2 8 561 1 8 561 227 0.65PUZZLE 9 214 1 650 17.9 7 564 10 1 379 2 ?ROTATE 7 504 719 9.6 6 785 18 5 662 198 0.69FREECELL 16 059 9 768 60.8 6 291 1 6 290 132 0.39
ICT4AgeingWell Conference, 20-22 May, 2015
11/19
Heavy noise present Outliers (usually caused by interrupts) are
omitted. Short term fluctuations (random differences
between consecutive puzzles, minor environmental disturbances, by tiredness, etc. )
Short-term fluctuations: zero-mean, stable independent random variables
Performance measured on a reference set will be compared to performance on the current set (no comparison is based on single game performance)
22 May, 2015
How to assess the cognitive state?
ICT4AgeingWell Conference, 20-22 May, 2015
12/19
Performance measure of a player during nearly one year Noisy – distribution of reference and current
data sets are to be compared
22 May, 2015
How to assess the cognitive state?
ICT4AgeingWell Conference, 20-22 May, 2015
13/19
Lilliefors test: the normality hypothesis is rejected
Nonparametric statistical hypothesis tests: Kolmogorov-Smirnoff two-sample test, Wilcoxon signed-rank test, etc.
22 May, 2015
How to assess the cognitive state?
ICT4AgeingWell Conference, 20-22 May, 2015
14/19
No direct proof - due to the long time needed to detect a critical cognitive change
Studies have shown that MCI patients performed poorly on Paired Associates Learning (PAL) computerized cognitive test. Players were asked to perform it.
The performance shown in the computer games correlates to their performance measured by the PAL test.
22 May, 2015
Proof of the concept
ICT4AgeingWell Conference, 20-22 May, 2015
22 May, 2015
15
M3W playground with games
16/19
22 May, 2015
Trail Making Test vs. Rabbits
17/19
Thank you for your attention!
22 May, 2015 ICT4AgeingWell Conference, 20-22 May, 2015
The M3W project has been supported by the AAL Joint Programme (AAL-2009-2-109), and by the national funding agencies in Hungary, Greece, Luxembourg and Switzerland.
Members of the project consortium:• Budapest University of Technology and Economics (HU)• Semmelweis University (of Medical Sciences, HU)• Zürich University of Applied Sciences (CH)• Actimage Ltd. (LU)• Silver Publishing LtD. (HU)• Gaudiopolis Elderly Home (HU)• Frontida Zois Homecare Ltd. (GR)https://m3w-
project.euhttps://kognito.eu