Hemelings R, Elen B, Barbosa-Breda J, et al. - In this investigation, researchers evaluated the use of deep learning (DL) for computer-assisted glaucoma identification, and the effect of training using images selected by an active learning strategy, which reduces labelling cost, as well as concentrated on the explainability of the glaucoma classifier. In order to develop a deep learning-based classifier for glaucoma diagnosis, this original investigation pooled 8433 retrospectively gathered and anonymized colour optic disc-centred fundus images. Based on optic disc-centred fundus images, the benefits of deep learning for automated glaucoma detection was demonstrated. In the medical community, the combined use of transfer and active learning can optimize DL model performance while minimizing the labeling cost of domain-specific mavens. Glaucoma specialists are able to use heat maps produced by the deep learning classifier to evaluate its decision, which appears to be linked to inferior and superior neuroretinal rim (within ONH), and RNFL in superotemporal and inferotemporal zones (outside ONH).