Impact of In-air Movements on Synthetic Data Generation for Handwriting Analysis in Neurodegenerative Diseases

The article discusses the impact of in-air movements on the generation of synthetic samples and classification tasks in Alzheimer’s disease (AD) diagnosis. The study used a Generator-Discriminator architecture to generate new samples and classify them into fake or real ones. The authors found that the increase in in-air time was due to the deficit in motor programming among AD patients, which could be a possible reason for their visuospatial deficit. The study also introduced a new dataset called DARWIN, which includes handwriting samples from 174 individuals: 89 AD patients and 85 healthy individuals. The dataset was obtained by writing with a pen on an A4 sheet of white paper placed over a Wacom’s Bamboo tablet, and the recorded information includes timestamp, x coordinate, y coordinate, binary pen-down property, and the pressure applied by the pen on the paper. The authors suggested that visuospatial deficit among AD patients could be a possible reason for their difficulty in performing everyday activities.
In simple terms, the article explores how people with Alzheimer’s disease (AD) move their hands when writing, and how this affects the accuracy of machines learning to recognize their handwriting. The researchers used a special machine learning model that can generate new samples and classify them as real or fake. They found that AD patients take longer time to start moving their hands, which might be because they have trouble with motor programming. The study also introduced a large dataset of handwriting samples from both healthy individuals and AD patients, which can be used to train machines learning models to recognize AD patient’s handwriting. Overall, the article highlights the importance of understanding how people with AD move their hands when writing to develop accurate machines learning models for diagnosing AD.