Use of the Clock Drawing Test and the Rey–Osterrieth Complex Figure Test-copy with convolutional neural networks to predict cognitive impairment

Abstract Background The Clock Drawing Test (CDT) and Rey–Osterrieth Complex Figure (RCFT) are widely used as a part of neuropsychological test batteries to assess cognitive function. Our objective was confirm the prediction accuracies RCFT-copy CDT for impairment (CI) using convolutional neural network algorithms screening tool. Methods data were obtained from patients aged 60–80 years who had more than 6 education. In total, 747 980 figures utilized. Convolutional TensorFlow (ver. 2.3.0) on Colab cloud platform ( www.colab.research.google.com ) preprocessing modeling. We measured accuracy each drawing 10 times this dataset with following classes: normal cognition (NC) vs. mildly impaired (MI), NC severely (SI), CI (MI + SI). Results better differentiating MI (CDT, 78.04 ± 2.75; RCFT-copy, not being trained) SI 91.45 0.83; 90.27 1.52); however, at predicting 77.37 1.77; RCFT, 83.52 1.41). 3-way classification (NC SI) approximately 71% both tests; no significant difference found between them. Conclusions two tests showed good performance severe cognition; alone is enough predict overall CI. There some limitations our study: sample size small, all participants did perform only copy condition RCFT used. Algorithms involving memory longitudinal changes worth future exploration. These results may contribute improved home-based healthcare delivery.