top of page
Abstract
Clustering is used in weakly and self-supervised learning to group similar images of tissue samples together. Unsupervised clustering allows for exploration of the latent space which is beneficial in digital pathology classification tasks. However, it is currently difficult to assess cluster quality in a quantitative manner for histology patches. Therefore, we propose a dimensionality reduction software pipeline. This would include an unsupervised clustering method pretrained on histology patches. Post clustering, the visualisation of the clusters is viewable from a GUI application that allows for interactive dimensionality reduction. This facilitates different analysis algorithms such as the Silhouette Coefficient and Dunn's Index. Furthermore, a qualitative assessment is easy by clicking on points in the graphical clusters and viewing the corresponding histology patch.
bottom of page