Fig. 3: The UMAP clustering result of Mueller matrix pixel images and their corresponding pathological features. | Communications Engineering

Fig. 3: The UMAP clustering result of Mueller matrix pixel images and their corresponding pathological features.

From: Unsupervised learning of pixel clustering in Mueller matrix images for mapping microstructural features in pathological tissues

Fig. 3

The six polarization feature clusters segment pathological features on the histological images (a), and pathologists identify them as cell nucleus (cluster 1 and 2), cytoplasm (cluster 3 and 4), and collagen fiber and fibrocyte (cluster 5 and 6). The heatmap of the Uniform Manifold Approximation and Projection (UMAP) is drawn at various hepatocellular carcinoma (HCC) differentiation degrees (b), to help visualize variations in polarization features corresponding to pathological variation. An animated version of (b) is supplied in Supplementary Video 1.

Back to article page