In generative AI models the image generation starts with a random seed, a so called latent. The latent contains information on the features in the final image e.g. the hair color of the person. For the random images on the right a random latent is chosen, while the average image is sampled by averaging all patients' features in a cohort.

personAverage Patient

personGestaltGAN Random

personGestaltDM Random

This patient does NOT exist

KBG