Description

Our red dot® algorithm is based on state-of-the-art deep learning models. Thanks to their specific architecture and ability to model complex functions, deep learning models have proved to be extremely efficient at processing images. Our red dot® algorithm was trained on over 30,000 CXRs with detailed annotations from certified radiologists. The training process teaches the algorithm to both classify a CXR and localize its findings. This localization is rendered through heatmaps and allows to understand and interpret red dot® output.