Multi-digit processing and contextualized analysis on the digital symbol digit test
Author(s)
DeTienne, Elizabeth A.
Download1192544154-MIT.pdf (3.539Mb)
Other Contributors
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science.
Advisor
Randall Davis and Dana L. Penney.
Terms of use
Metadata
Show full item recordAbstract
Neurocognitive decline has been shown to occur as early as 15-20 years before obvious symptoms develop for patients with Alzheimer's. Current therapies work best when begun early, but it is very difficult to detect subtle cognitive change before obvious symptoms manifest. We have done analysis on the handwritten data from the digital Symbol Digit Test with the aim of detecting subtle cognitive decline early in the disease progression. We used a large dataset to augment the MNIST handwritten digit dataset. This has enabled us to recognize digits 0-12 with high accuracy, making it possible to automate scoring of the test. We also analyzed subtle features of the handwriting. We contextualized this data through visualizations, revealing a number of interesting trends and deviations for healthy patients versus patients with cognitive decline. For example, impaired participants tend to have more ink than we would expect for their average digit height, and pause for longer before writing a digit. We believe that this analysis will provide valuable new insights into a person's cognitive status.
Description
Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, May, 2020 Cataloged from the official PDF of thesis. Includes bibliographical references (pages 53-55).
Date issued
2020Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer SciencePublisher
Massachusetts Institute of Technology
Keywords
Electrical Engineering and Computer Science.