Alyssa Lanzi
Communication Sciences and Disorders
University of Delaware
lanzia@udel.edu
|
|
Alyssa Lanzi Communication Sciences and Disorders University of Delaware lanzia@udel.edu |
| Participants: | 184 MCI and 299 neurotypical, data collection ongoing |
| Type of Study: | Longitudinal discourse study |
| Location: | USA |
| Media type: | audio |
| DOI: | doi:10.21415/Q0JX-5W20 |
In accordance with TalkBank rules, any use of data from this corpus must be accompanied by at least one corpus reference.
Lanzi, A. M., Saylor, A. K., Fromm, D., Liu, H., MacWhinney, B., & Cohen, M. (2023). DementiaBank: Theoretical rationale, protocol, and illustrative analyses. American Journal of Speech-Language Pathology, 32(2), 426-438. https://doi.org/10.1044/2022_AJSLP-22-00281
When publishing articles or reports related to this data set, the following language is also suggested to acknowledge federal funding supporting this work:
Data collection was supported by the National Institute of Aging of the National Institutes of Health under award number RF1AG083823 (PIs: MacWhinney and Lanzi).
Click on these links below for more information about dementia, DementiaBank Grand Rounds, the DementiaBank website, and the Delaware corpus.
Understanding the difference between Alzheimer's disease and dementia
Learn more through DementiaBank Grand Rounds
Note: Access to the video clips in Grand Rounds requires registration and approval through DementiaBank. Please see the information at the top of the DementiaBank website here for instructions on how to register.
Overview of DementiaBank Website and the Development of the Delaware Corpus
This work was supported by the National Institute on Deafness and Other Communication Disorders under grant R01DC008524-13S1 awarded to Brian MacWhinney.
Overview of Data Collection Process
Overview of Cognitive Classification Process
Overview of Demographic Spreadsheet
The DementiaBank Delaware Corpus discourse protocol is administered to participants with mild cognitive impairment (MCI) and those without cognitive impairments. The protocol includes:
Directions for administering the discourse protocol.
*NOTE: The standard administration for this Cinderella story narrative task involves having the Participant first look through pictures from the Cinderella storybook to
refresh their memory of the story. The pictures are then removed before
the Participant tells the story. In some cases, however,the Participant
was given the option of looking through the pictures and chose not to.
These are the Participants who did NOT look through the Cinderella
pictures prior to telling the story:
MCI: 5, 10, 14, 15, 20, 21, 27, 28, 31, 33, 40, 43
Neurotypical: 3, 8, 13, 24, 30, 34, 35, 36, 38, 39
Cognitive Linguistic testing includes:
1a. Montreal Cognitive Assessment (MoCA v.7.1)
Nasreddine, Z. S., Phillips, N. A., Bédirian, V., Charbonneau, S., Whitehead, V., Collin, I., Cummings, J. L., & Chertkow, H. (2005). The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. Journal of the American Geriatrics Society, 53(4), 695–699.
A 30-question cognitive screening test designed to detect mild cognitive impairment and Alzheimer's disease.
PDF here.
OR
1b. MoCA Blind/Telephone MoCA
Dawes, P., Pye, A., Reeves, D., Yeung, W. K., Sheikh, S., Thodi, C., ... & Leroi, I. (2019). Protocol for the development of versions of the Montreal Cognitive Assessment (MoCA) for people with hearing or vision impairment. BMJ open, 9(3), e026246.
PDF here.
2a. Boston Naming Test-Short Form (BNT-SF; Kaplan, Goodglass, & Weintraub, 2001)
OR
2b. Multilingual Naming Test
Stasenko, A., Jacobs, D. M., Salmon, D. P., & Gollan, T. H. (2019). The Multilingual Naming Test (MINT) as a measure of picture naming ability in Alzheimer’s disease. Journal of the International Neuropsychological Society, 25(8), 821-833.
PDF here.
Gollan, T. H., Weissberger, G. H., Runnqvist, E., Montoya, R. I., & Cera, C. M. (2012). Self-ratings of spoken language dominance: A Multilingual Naming Test (MINT) and preliminary norms for young and aging Spanish–English bilinguals. Bilingualism: Language and Cognition, 15(3), 594-615.
PDF here.
3. Hopkins Verbal Learning Test – Revised (HVLT-R)
Benedict, R. H. B., Schretlen, D., Groninger, L., & Brandt, J. (1998). Hopkins Verbal Learning Test–Revised: Normative data and analysis of inter-form and test–retest reliability. The Clinical Neuropsychologist, 12(1), 43–55. https://pmc.ncbi.nlm.nih.gov/articles/PMC6757330/
4. Wechsler Memory Scale – Revised: Logical Memory I and II (WMS-R)
Wechsler, D. (1987). Manual for the Wechsler Memory Scale-Revised. The Psychological Corporation, San Antonio, TX.
A measure of episodic memory through immediate and delayed short story retell.
More information about PROTOCOL MATERIALS can be found here.
Other Info -- Transcripts and Demographics
The CHAT transcript ID header tier for the Participant includes the following information:
NOTE: Please be aware that some transcripts in the Download transcripts folder are still being reviewed for accuracy after being generated by ASR. The transcripts still under review have this Comment line in the header tiers: Unchecked output of ASR model; do not use. These transcripts will not have %mor, %gra, or %wor tiers until they are fully reviewed, at which time the Comment line will also be removed.
Complete demographic and test data are available in a spreadsheet in the downloadable transcript folder.
The PREPARE Challenge, sponsored by the National Institute on Aging with support from NASA, was designed to advance innovative and representative early prediction of AD/ADRD through a multi-phase approach emphasizing both methodological rigor and equity.
Phase 1. An initial phase of the challenge focused on identifying open datasets capable of supporting machine learning approaches while explicitly addressing data biases in dementia research. Among diverse submissions spanning behavioral assessments, medical records, neuroimaging, and less conventional sources, the first-place team from SUNY Buffalo demonstrated the potential of acoustic analysis using features extracted from the Delaware corpus and other data within DementiaBank. This winning submission was selected based on the utility and rigor of target variables and predictors, innovation and field-advancing potential, capacity to address existing research and diagnostic biases, and practical usability.
Phase 2. DementiaBank emerged as one of only two datasets from Phase 1 deemed sufficiently comprehensive and well-documented to support an open machine learning competition. In Phase 2, it anchored a data track that attracted 401 participants who developed innovative models for early detection of AD/ADRD using audio data using a subset of DementiaBank that was modified to best support a competition. Competitors were challenged to improve early prediction accuracy using voice-based features potentially indicative of cognitive impairment, but were also encouraged to explore linguistic patterns and semantic content.
Phase 3. In a final challenge Phase, winning teams spent several months refining their Phase 2 models and incorporating additional data, representing a shift from pure performance optimization to practical implementation considerations. While Phase 2 prioritized performance metrics, Phase 3 emphasized making models more fair, equitable, and generalizable across diverse populations. Top teams in Phase 3 took advantage of DementiaBank's full breadth, benchmarking performance across corpora and improving models by training with complementary external datasets like Mozilla Common Voice.
You can learn more about the overall Challenge and its results here.
The data can be accessed here.
Note: Access to the data requires registration and approval through DementiaBank. Please see the information at the top of the DementiaBank website here for instructions on how to register.
Additional information about the dataset and metadata is available here.