Development and validation of a dementia prediction model by analysing digital drawing behaviour
Refereed conference paper presented and published in conference proceedings

替代計量分析
.

其它資訊
摘要Background
Traditional dementia screening tools are mainly administrated in pencil-and-paper form, and the result is totally relied on interviewers’ subjective judgement. A digital platform captures real-time temporal and spatial drawing image, and to recognize the image’s features (1). Pilot study showed that using the drawing behaviors and image features can help to improve the screening accuracy (2). The aim of this study is to validate a digital drawing platform with participants with or without clinical diagnosed Alzheimer’s’ disease (AD).

Methods
Clinical diagnosed AD patients were recruited in a geriatric research clinic (GRC). Healthy controls (HC) score 22 or above in Montreal Cognitive Assessment (MoCA) test were recruited from an osteoporosis study, a GRC and seven community centers in Hong Kong. MoCA was administrated before the participants draw two interlocking pentagons on the digital drawing platform. Participants were randomly assigned to the derivation and validation cohorts in 2:1 ratio. Drawing behaviors and image features were constructed in the derivation cohort. Multivariate logistic regression was used to generate the prediction model. Drawing behaviors included time of drawing the image, time stopped for thinking, and speed of image drawn. Image features included number of angles drawn in the pentagons, presence of overlapped pentagons etc.

Results
A total of 465 participants were recruited, and 194 of them (41.72%) were AD patients. The mean age (SD) of AD patients and HC were 79.6 (6.2) years and 75.6 (7.7) years respectively. In the derivation cohort, 118 out of 310 participants were AD diagnosed with mean (SD) MoCA score of 11.9 (4.6), and 192 participants were HC with mean (SD) MoCA score of 24.9 (2.1). Regression model showed image features, such as non-overlapping pentagons (OR, 95%CI: 2.69, 1.11-6.54), and drawing features, such as increase in maximum speed of drawing (OR, 95%CI: 1.04, 1.02-1.07), contributed to higher risk of dementia (Table 1). The Area Under Curve (AUC) of Receiver Operating Characteristic (ROC) curve was 0.72 from the validation cohort.

Conclusions
Drawing behaviors captured by digital platform improves the accuracy of dementia screening. In the future, technology assisted platform will be an important step for self-monitoring on cognitive function.
著者Wong M, Lam MWY, Wong A, Kwok T, Wong SYS, Tsoi K
會議名稱Alzheimer’s Association International Conference
會議開始日22.07.2018
會議完結日26.07.2018
會議地點Chicago
會議國家/地區美國
會議論文集題名Alzheimer's & Dementia
系列標題Poster Presentations: Monday, July 23, 2018
叢書冊次P2-336
出版年份2018
月份7
卷號14
期次Suppl 7
出版社Elsevier
頁次812 - 812
國際標準期刊號1552-5260
語言英式英語

上次更新時間 2020-18-09 於 02:12