JMIR Publications

ABSTRACT

Background:

Alzheimer's disease (AD) and other types of dementia is now one of the world's worsening health problems for aging people and the fifth dominant cause of death over the world. Due to the growing population with dementia and the increasing cost of dementia, early detection of the disease at the stage of mild cognitive impairment (MCI), which is a prodromal stage of progressing to AD and mild AD, is beneficial to improve the life quality of patients and to reduce the loading of their caregivers.

Objective:

This study aimed to design a digital screening system according to the Rey-Osterrieth Complex Figure (ROCF) neuropsychological test in order to assist the clinicians to assess whether the participant is MCI or AD against health control (HC) automatically.

Methods:

Our system utilized the data-driven deep learning approach through a convolutional neural network which was designed for pre-training and extracting useful features from an open sketch dataset. The learned features were then transferred to our collected dataset for further training of the classifier. We evaluated our performance using area under the receiver operating characteristic (AUC).

Results:

An average area under the receiver operating characteristic curve score (AUC) of 0.913 was achieved for classifying MCI vs. HC in the traditional pencil and paper dataset. On the other hand, a dataset that was collected using digitalize graphics tablet and smart pen achieved an average AUC of 0.950 in classifying AD vs. HC.

Conclusions:

We proposed a digital screening system is able to distinguish MCI or AD participants from HC on the basis of the Rey-Osterrieth Complex Figure neuropsychological test. The collected hand-drawn images from the participants are discriminated using deep learning and transfer learning techniques. The proposed system can be used to assist the clinicians in early diagnosis the conditions of the patients and reduce the burden of them.