Figure 7.

Box plots for the AUC across 50 runs from all 3 algorithms (Φ_ms, Φ_t, Φ_b). The red line identifies the mean, the blue box encompasses 25^th percentile, with the black whiskers extending to the 75^th percentile. Red dots are indicative of outliers. We can see that the MS provides a higher mean AUC compared with the texture features, resulting in a significantly a smaller variance. The homogeneity constraint imposed by ball-scale does not appear to provide discrimination between TILs and non-TILs