Table 3.

Comparison of C-index of the cox regression model, -log10(p-value) of survival analysis, the number of prognosis-related features, and the percentage of p-value less than 0.05 on the training and testing set in cross-validation between the prognostic prediction model using morphological features extracted from cells in the tumor region and the model using features in regions randomly selected from the whole slide.

Region	C-index of cox regression	-log10(P-value of KM estimate)	Number of prognosis-related features	Number of prognostically significant features	Percentage of P-value less than 0.05 on the training set	Average p-value on the testing set
Tumor	0.7007	5.825	11	6	100%	3%
Randomly selected	0.6637	4.191	5	3	92%	7%