Analysing Dynamic Data for MoCA Solo #7
Description
Title
Analysing Dynamic Data for MoCA Solo
Leaders
Murray Gilles: murray.gillies@mocacognition.com
Saber Naderi: saber.naderi@mocacognition.com / Discord handle: saber_moca
Yijae Kim: yijae.kim@mocacognition.com / Discord handle: yijae1_42148_49872
Collaborators
Émile Dusablon Dion: emile.dusablondion@mocacognition.com / Discord handle: emile_98830
Jack Tacchi: jack.tacchi@osedea.com / Discord handle: jackosedea
Walid Masoudimansour: walid.masoudimansour@osedea.com / Discord handle: walid.osedea
Project Description
MoCA Cognition has developed a new digital tool called MoCA Solo to quantify a patient’s cognitive performance. It is based on the very well-established paper test “MoCA 8.1”, used to assess people with Mild Cognitive Impairment and Alzheimer’s disease. It requires the patient to perform an array of tasks, such as clock drawing, naming animals or recalling 5 words. For the paper version, a point is awarded for each task, but how the task is completed can only be observed live by a human. In the MoCA Solo application, all data is recorded in raw format, i.e. all audio files and every click on the iPad are collected for post data collection analysis.
MoCA Cognition has collected data with the MoCA Solo application from 500 English-speaking participants, including patients and healthy individuals. All the current outcome measures have been annotated by three human raters. MoCA Cognition has used this data to develop AI scoring algorithms to automatically score the current outcome measures without the need for the presence of a human. These scores characterize the patient and the goal of this activity was to reproduce the paper MoCA score.
While the current outcome measures are available from both the algorithms and the ground truths for the 500 people, we haven’t explored the dynamics of how the MoCA Solo tests were completed. We are interested in whether there is information in the dynamics of how the tests are completed which correlates with the existing outcome measures. An example of this could be the time it takes for a patient to name an animal shown in an image and the ability to recall the 5 words in delayed recall. The goal of the project is to create algorithms that take the dynamic data and find features that correlate with the existing measures such as the MoCA Score.
Demo video of MoCA Solo: https://youtu.be/sl_4nr-n8SM
MoCA Paper 8.1 & Instructions : https://captiva.neurosurgery.ufl.edu/resources/moca/
Link to project repository/sources
No response
Goals for Brainhack Montreal
The main goal of this project is to create algorithms that take the dynamic data from MoCA Solo and find features that correlate with the existing measures such as the MoCA Score.
Sub-goals:
-
Create an algorithm that measures the Naming reaction times (as accurately as the ones manually annotated in the reference dataset called MCTOR_ANN).
-
Apply the algorithm to the unannotated data called MCTOR_UNA and check manually to see if the algorithm performance is good.
-
Use the reaction times to predict delayed recall score.
-
Find the predictive power of the reaction times in predicting whether the MoCA score of a participant is above or below 25.
-
Repeat this reaction time analysis for the immediate memory, delayed recall audio task and try to find the predictive power of the combined reactions times (delayed recall reaction times and Naming reactions times). To able to do this, manual annotation of delayed recall audio will be required (consider starting this at the beginning of the project, to have enough time to annotate).
Skills
- Data science & analysis
- Programming (Python or R)
- Machine learning / algorithm development
- Neuropsychology / Cognitive science (not mandatory, but nice to have)
Tech stack
- Python Programming language
- Python data libraries
- Machine learning libraries (scikit-learn, etc.)
Onboarding documentation / Expectations
No response
What will participants learn?
- Feature engineering from dynamic cognitive task data
- Correlation analysis
- Algorithm development
- Interdiciplinary thinking
- Collaboration in a real research workflow
Data to use
Those who would like to work on the MoCA Solo data will be granted access through SharePoint.
Credit to collaborators
No response
Image
No response
Type
coding_methods
Project Maturity Status
2 - releases existing
Topic
statistical_modelling
Tools
Jupyter
Programming language
Python
Modalities
not_applicable
Git skills
1_commit_push
Anything else?
No response
Things to do after the project is submitted and ready to review.
- Short summary of your project pitch to present during the hackathon.