A novel deep learning-based algorithm combining histopathological features with tissue areas to predict colorectal cancer survival from whole-slide images

Table 1 Performance of the histopathological features via fivefold cross-validation using C-index values

Method	TUM^a	LYM^a	STR^a	MUC^a	TUM + LYM + STR + MUC^a
LASSO-Cox	0.521 ± 0.041	0.448 ± 0.054	0.566 ± 0.042	0.559 ± 0.065	0.687 ± 0.084
RIDGE-Cox	0.615 ± 0.059	0.567 ± 0.068	0.532 ± 0.107	0.589 ± 0.058	0.616 ± 0.092
EN-Cox	0.546 ± 0.036	0.443 ± 0.065	0.539 ± 0.037	0.573 ± 0.079	0.646 ± 0.064
SSVM	0.616 ± 0.042	0.565 ± 0.069	0.479 ± 0.085	0.596 ± 0.088	0.598 ± 0.095
RSF	0.601 ± 0.057	0.455 ± 0.065	0.498 ± 0.111	0.556 ± 0.107	0.605 ± 0.078
GBRT	0.551 ± 0.087	0.443 ± 0.036	0.536 ± 0.062	0.560 ± 0.058	0.610 ± 0.052

The results highlighted in bold black show the best performance with those methods
^aLYM lymphocyte, MUC mucus, STR stroma, TUM tumor

ISSN: 1479-5876