Can clustering of deep image representations reveal latent breast cancer risk subtypes in a multiethnic population?
Date:
Sollis L, Wolfgruber T, Quon B, Kane C, Valdez D, Hernandez BY, Kerlikowske K, Sadowski P, Shepherd JA.
1,283 cases and 3,159 matched controls; six clusters derived from deep representations differed significantly across nearly all key clinical variables (Bonferroni-adjusted p < 0.008).