Comparison of transfer learning performance metrics for models pretrained on baseline synthetic data derived from the dog spinal cord (baseline) and synthetic data with axon diameters scaled down by a factor of 0.15 (shrunk axons).

	Model
Metric	Baseline	Shrunk axons
Genu (% error in mean)	4.32% $\pm$ 4.86%	3.62% $\pm$ 3.97%
Body (% error in mean)	2.79% $\pm$ 2.73%	3.43% $\pm$ 3.36%
Splenium (% error in mean)	3.02% $\pm$ 2.30%	4.35% $\pm$ 3.56 %
Genu effect size	1.76 $\pm$ 1.23	1.29 $\pm$ 1.08
Body effect size	3.02 $\pm$ 3.19	1.55 $\pm$ 1.68
Splenium effect size	-0.39 $\pm$ -0.29	0.06 $\pm$ 0.07