Atlas H&E-TME

Atlas H&E-TME was validated against board-certified pathologist annotations across datasets from multiple sources and scanner types. All OpenTME outputs were produced using Atlas H&E-TME, an AI application built on Atlas pathology foundation models. Atlas H&E-TME applies a sequential four-stage workflow. The Atlas H&E-TME foundation model is a vision transformer pre-trained on diverse histopathology data using self-supervised DINO learning. Atlas H&E-TME uses supervised fine-tuning of the Atlas backbone with pathologist-informed annotations to create task-specific models.