Authors

Sarah Anderson MSc (1)
Heather Jamniczky PhD (1)
Olave Krigolson PhD (2)
Kent Hecker PhD (1)

Institutions

(1) University of Calgary - Alberta - Canada

(2) University of Victoria - British Columbia - Canada

Logo
  • Currently 5.00/5
  • 1
  • 1
  • 2
  • 2
  • 3
  • 3
  • 4
  • 4
  • 5
  • 5

Rating: 5.0/5 (2 votes cast)

Theme

5BB Team Based Learning/Learning anatomy

Title

Understanding 2D anatomy learning- a reinforcement task based approach

Background

  • Given reduced formal instruction time for many of the basic sciences within medical curricula, educators are searching for efficient instructional methods that ensure students have the necessary foundational knowledge. 
  • Reinforcement learning tasks may be an effective means to promote learning of base knowledge prior to classroom interactions in anatomy education. 

Summary of Work

Objective

  • To design a reinforcement learning task in which novice participants successfully learn to identify neuroanatomical structures in a time efficient manner
  • We predicted that provision of immediate feedback would activate reinforcement learning mechanisms within the brain thus enhancing knowledge acquisition such that performance accuracy (correct identification of neuroanatomical structures) improves from approximately 50% (guessing) to 90% by task completion

Participants

  • n = 10 recruited from health-related programs at the University of Calgary, Canada
  • Minimal neuroanatomical knowledge

Experimental Set-up

  • Task framework modified from Krigolson et al. 1
  • Task presented on a computer screen using a customized MATLAB program in conjunction with Psychophysics Toolbox extensions 2,3
  • Data tables generated for participant accuracy and response time

Anatomical Structure Identification Task:

  • Participants learned to identify 10 neuroanatiomical structures (2D coronal brain images) using positive and negative feedback
  • 16 blocks x 20 trials/block = 320 trials total

 

Figure 1. Information on learning task. 

A) Examples of 2D coronal brain sections, arrows indicate structure student learned to identify; B) List of structures students learned to identify; C) Generalized example of a trial. 

Summary of Results

  • Learning curves show learning occurs over the course of a task (320 trials) including 16 trial blocks (20 trials/block)
  • Participants consistently exceed 90% mean accuracy in block 13 (260 trials)
  • The total task duration was approximately 30-35 minutes with all participants reaching proficiency by 25-30 minutes
  • Significant increase in performance on a post knowledge test compared to a pre-test, M = 90.00% CI [81.57, 98.43], t(9) = 24.15, p < .001

 

Figure 2. Changes in performance accuracy for each participant (n = 10)

 

Figure 3. Comparison of performance accuracy on trials using correctly versus incorrectly labeled neuroanatomical images (±SEM; n = 10)

 

Figure 4. Response time for a trial during each block of the experiment (mean ± SD; n = 10)

Conclusion

  • Our results highlight the key role of reinforcement learning approaches to establishing foundational knowledge in the pre-clinical sciences, specifically anatomy
  • Instructors can assess student progression of learning through examination of learning curves
  • Future work will assess neurophysiological responses through measurement of event-related brain potentials using electroencephalography

Take-home Messages

  • Designing effective pre-class exercises that make use of reinforcement learning theory as a means to promote learning may be an effective method to build base knowledge prior to classroom interactions in anatomy education

Acknowledgement

We are thankful for the funding provided by the following sources:

University Research Grants Committee Seed Grant- University of Calgary

Queen Elizabeth II Graduate Doctoral Scholarship

Cumming School of Medicine, University of Calgary

Faculty of Veterinary Medicine, University of Calgary

Faculty of Graduate Studies Travel Award

References

  1. Krigolson OE, Pierce LJ, Holroyd CB, Tanaka JW. Learning to Become an Expert: Reinforcement Learning and the Acquisition of Perceptual Expertise. Journal of Cognitive Neuroscience. 2009;21(9):1833-40.
  2. Brainard DH. The psychophysics toolbox. Spatial vision. 1997;10:433-6.
  3. Pelli DG. The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial vision. 1997;10(4):437-42.
Background
Summary of Work
Summary of Results

Figure 2: Changes in performance accuracy for each participant (n = 10)

  • Repeated Measures ANOVA: significant effect of blocknumber on structure identification performance, F(15, 135) = 27.18, p < .001, partial eta squared = 0.75
    • Performance significantly improves from Block 1 to Block 4
    • Participants consistently exceed 90% accuracy in block 13 (260 trials)

 

Figure 3: Comparison of performance accuracy on trials using correctly versus incorrectly labeled neuroanatomical images (±SEM; n = 10)

  • Generally, mean accuracy was higher on correctly matched image-label pairings (M = 88.16%, SD = 13.94) compared to incorrect pairings (M = 83.75%, SD = 14.39) throughout the task (M = 4.41%, 95% CI [2.09, 6.73], t(15) = 4.05, p < .001)

 

Figure 4: Response time for a trial during each block of the experiment (mean ± SD; n = 10)

  • Repeated Measures ANOVA: significant effect of block number on response time, F(15, 2985) = 72.22, p < .001, partial eta squared = 0.27
    • Mean response time significantly improves from the first block (M = 1345 ms, SD = 386 ms) and fourth block (M = 1141ms, SD = 284 ms)
    • Mean response time plateaus in the last four blocks and did not significantly differ from the last block (M = 686 ms, SD = 283ms)
Conclusion
Take-home Messages
Acknowledgement
References
Send ePoster Link