Deep Learning Using Havrda-Charvat Entropy for Classification of Pulmonary Optical Endomicroscopy

•The use of Havrda-Charvat in loss function generalizes the classical Shannon based cross-entropy.•Our method is promising for small annoted databases.•Havrda-Charvat is a parameterized entropy and one can choose the parameter to fit the data.•Our method aims to separate uninformative optical images from informative ones.