Prognosis of Dementia Using Early Fusion Approach with Digital Clock Drawing and Trail-Making Tests

Abstract

Dementia poses a substantial global public health challenge. Emerging evidence suggests that COVID-19's neurological impact may aggravate dementia incidences. Timely recognition and management can significantly decelerate dementia progression and enhance affected individuals' quality of life. In the realm of cognitive assessment, the Clock Drawing Test (CDT) and Trail-Making Test (TMT) stand as prominent tools. Existing research predominantly focuses on the use of these tests in isolation. As CDT checks visuospatial skills and planning, TMT focuses on processing speed and mental flexibility. Combining these allows us to better understand an individual's cognitive strengths and weaknesses. This study aims to assess whether combining the features from the digital versions of these tests as dCDT and dTMT can enhance classification accuracy and recall in dementia cases. It utilizes an early fusion technique, merging feature metrics from both these tests for dementia classification. The study includes 86 healthy control participants and 52 individuals diagnosed with dementia. The early fusion method demonstrates promising outcomes as an alternative to conventional paper-based screening methods of CDT and TMT. The model attains a prominent overall accuracy of 93%, along with 87% precision, 85% recall and 0.94 AUC. The results exhibit reasonable improvements in classification performance as against prior studies involving individual modes of dCDT and dTMT. With the increase in the dataset size, this study can be extended for the classification of dementia sub-types. The scope of the study and data collection process is reviewed and approved by an independent ethics committee.