Fig. 4
From: Integrating holotomography and deep learning for rapid detection of NPM1 mutations in AML
Deep learning classification of myeloblast holotomography (HT) images for NPM1 mutation status. (a) Validation result of the deep learning classifier on myeloblast HT images of four NPM1 wild-type patients. (b) Validation result of the deep learning classifier on myeloblast HT images of four NPM1 mutant patients. (c) Receiver operating characteristic curve representing the performance of patient-wise NPM1 mutation screening based on different threshold positive prediction ratios. (d) Plot of deep learning output values corresponding to the two classes (wild-type and mutant). Each data point represents an individual myeloblast HT image. The dashed diagonal line indicates the boundary where the two output values are equal.
