Investigating the Effects of Model Size on Anatomy Learning

Abstract Body : Introduction: Compared to cadaveric specimens, three dimensional (3D)-printed anatomical learning objects often offer lower production costs, easier storage, and greater durability as well as a higher level of customization. This production flexibility leads to a question that was previously unnecessary to ask since anatomical objects came only in their native size. What model size is best for learning? Research into manipulating different models to optimize anatomy learning is non-existant. This study investigates the effect of model size on students’ ability to learn anatomical landmarks as well their qualitative feedback on model features, including label size, model color, and handling ability.

Methods: The study population consisted of 380 undergraduate students with no prior anatomical knowledge. Participants were randomized to learn from either vertebra and pelvic bone models (os coxae), or scapula and sphenoid bone models. The size of the 3D-prints that they received was also randomized. Bone models were 3D-printed at 50%, 100%, 200%, and either 300% or 400% of their reference specimen size. Participants were tested on nominal anatomy recall immediately after learning from each of their two assigned models. Mental rotation ability and working memory were also assessed, and opinions regarding learning with the various models were solicited.

Results: Model size was calculated by measuring the longest diameter. This was found to have a small, but significant effect on test score (F(2,707) = 17.15, P < 0.05, R2 = 0.046). Moreover, mean test scores for participants studying from any model with a longest diameter above 10 cm were greater than 75% and were significantly higher than those with longest diameter below 10 cm (Kruskal-Wallis H Test, χ2 (1) = 20.232, P < 0.001). All anatomical objects were not equally challenging to learn from. Notably, average test scores for participants studying scapula models were significantly higher when compared to every other model (P ≤ 0.001), with all scapula sizes yielding mean scores greater than 85%. There were no significant differences (P = 0.758) between test scores of participants using scapula models with longest diameter above and below 10 cm.

Linear regression analysis demonstrated moderate positive correlation (R = 0.324) between MRT scores and test scores, while a slightly weaker positive correlation was observed between OSPAN and test scores (R = 0.282). Both correlations were statistically significant (P < 0.001).

For features impacted by model size (i.e., label size and handling ability) qualitative analysis indicates negative opinions were associated with very small and very large models. Intermediate sizes generated mostly positive feedback. Interestingly, the scapula was associated with the most positive comments, regardless of size.

Conclusion: This study suggests that, regardless of the native size of the anatomical object, a size near 10 cm optimizes learning and minimizes negative feedback about size and size-related features.

Translating its findings commercially, models need not be produced with a longest diameter of > 10 cm which will save on production costs (e.g., expensive 3D-printing filaments and time) without a loss of benefit to learning. As this study investigated bones, which are a uniformly hard material and easily replicable as 3D models, one area of further investigation could be to examine the importance of material realism when learning from 3D models that represent softer tissues.