Cognitector2024
Description
Cognitector2024 This dataset was curated for monitoring attention, executive function and other cognitive states among university students using the trail-making test (TMT-A and TMT-B) experiment as a PhD research project to build an interpretable model based on a vision transformer for detecting such cognitive states based on facial expressions in a human video sequence. Forty-three video recordings of students who took the TMT-A and TMT-B test were originally in a .mov format, but were converted to .mp4 for the deep learning model. Each video duration spans from 13 seconds to a maximum of 60 seconds. By convention, an individual who completes the TMT-A test in less than 25 seconds was classified as having a Fast Attention Rate (FAR), while individuals who took more than 25 to 60 seconds to complete the task were considered to have an Average Attention Rate (AAR). The 25-second decision boundary for the class “Fast” against “Average” attentions is supported by the meta-analytic norms according to Tombaugh [1], where durations ≤ 25 seconds represent above-average performance (in the top 30%). Lezak et al. [2] clinical quartile splits and cognitive correlates [3] further support this, where a sub-25-sec performance reflects superior visual attention and processing speed. Thus, a total of 22 students were classified as AAR, while 21 were considered FAR. One outlier was observed in the AAR class, where the last sample shows a rather wider gap of over 10 seconds; otherwise, the time difference between the two classes was evenly distributed. Each video dataset consisted of 59.94 frames per second (fps), with each frame having a width of 1280 and a height of 720. The dataset on the TMT-B was not included in this version as it has not yet been tested. References [1] T. N. Tombaugh, “Normative Data for the Trail Making Test (TMT) in Healthy Adult Populations: A Meta-Analysis,” J Clin Exp Neuropsychol, vol. 26, no. 2, pp. 159–178, 2004, doi: 10.1076/jcen.26.2.159.28087. [2] M. D. Lezak, D. B. Howieson, E. D. Bigler, and D. Tranel, Neuropsychological Assessment, 5th ed. New York, NY: Oxford University Press, 2012. [Online]. Available: https://global.oup.com/academic/product/neuropsychological-assessment-9780195395525 [3] I. Sánchez-Cubillo et al., “Construct validity of the Trail Making Test: Role of task-switching, working memory, inhibition/interference control, and visuomotor abilities,” 2009, Cambridge University Press. doi: 10.1017/S1355617709090626.
Files
Steps to reproduce
1. Prepare participants to understand the task by watching instructions or explaining the purpose of the cognitive task 2. Step up a touchscreen-enabled computer with the computerized version of the trail-making test (TMT-A and TMT-B) experiments in a browser. (This is useful to allow participants to participate without restrictions and delays with the use of mouse) 3. Step a high definition camera positioned to capture the full human face before the participant starts the task. 4. Ask the participant to start the task, while the camera starts recording the simultaneously records the facial expression as they undertake the task. 5. Pause the recording after the first task for the TMT-A is done. (Video is automatically saved). 6. Record the age, gender and corresponding time and score of the individual. 7. Ask participant to commence the second part of the task (TMT-B) while you record. 8. Repeat Steps 5 and 6. 9. Save all recorded video and use the clinically recommended threshold to classify the videos based on duration.
Institutions
- Lancaster University - Ghana Campus
- Kwame Nkrumah University of Science and Technology College of Science