The digital twin of the human brain: Simulation and assimilation

Abstract The Digital Twin Brain (DTB) is a simulation of the human brain at the large scale, with up to 86 billion neurons and 47.8 trillion neural synapses. The DTB mimics both the resting state and aspects of the action of its biological counterpart. A novel architectural assignment of neurons to 14,012 GPUs (with 57.39 million cores) and a two-level routing scheme between GPUs was implemented in the simulations. This was combined with development of a hierarchical mesoscale data assimilation method, that is capable of constructing trillions of parameters from estimated hyper-parameters. The constructed DTB reproduces blood-oxygen-level-dependent (BOLD) signals of the resting-state of its living biological counterpart with high correlation (> 0.9). We enabled the DTB to interact with its environment, as demonstrated in a visual task. The result indicates the feasibility of implementing a digital twin of the human brain, by overcoming key computational, mathematical and biological challenges, and opens the door to a broad range of applications.