Fig. 2: Representative patches of machine-learned features associated with lymph node metastasis. | Communications Medicine

Fig. 2: Representative patches of machine-learned features associated with lymph node metastasis.

From: Predicting lymph node metastasis from primary tumor histology and clinicopathologic factors in colorectal cancer using deep learning

Fig. 2: Representative patches of machine-learned features associated with lymph node metastasis.

Univariable AUROCs, univariable odds ratios (OR) and sample patches for the top 5 machine-learned features are shown. Patches representing the same machine-learned features are shown for the temporal validation set and the external validation set 1a, respectively. Patches selected here are closest to each cluster centroid, and each patch within a single feature is sampled from a unique case. Patches are 289 × 289 pixels obtained at 10X, with scale bar in lower right showing length of 100 micrometers.

Back to article page