Table 3 Spot-level performances of the weakly supervised models trained with or without knowledge distillation, on the PETACC8-Visium test set of 24 samples

From: A deep learning-based multiscale integration of spatial omics with tumor morphology

Distillation

Pearson correlation

Spearman correlation

Without

0.369 [0.308; 0.425]

0.371 [0.301; 0.436]

With

0.427 [0.355; 0.491]

0.430 [0.346; 0.502]

  1. 95% confidence intervals were obtained by bootstrapping samples 10,000 times with replacement.
Back to article page