Towards intuitive simultaneous control of a bionic limb

Restoring arm function after limb loss with a prosthesis remains a major challenge. Recent advances in surgical techniques and engineering approaches are now enabling substantial restoration of functionality after amputation. This doctoral thesis investigates cutting-edge surgical and engineering strategies and their integration, aiming to achieve intuitive, simultaneous control over multiple bionic joints in myoelectric prostheses, thereby surpassing current clinical solutions.A key focus was to understand how residual biological pathways after amputation, which naturally encode volitional movement, can be harnessed. We demonstrated that severed nerves can be redirected to innervate denervated native muscles and free muscle grafts, creating new, long-term stable myoelectric sources. These enabled simultaneous, proportional control of up to three degrees of freedom using a conventional one-to-one mapping strategy, improving functionality and reducing disability during extended home use. To further enhance motion-intent decoding and increase the number of controllable boinic joints, we explored deep learning methods and biologically inspired data-collection techniques for training neural networks. Our results show that deep learning architectures outperform shallow networks, facilitating intuitive simultaneous control. We further demonstrated that artificial training data can greatly reduce the burden of lengthy fitting sessions. These methods enabled intuitive, simultaneous, proportional control over 4.5 degrees of freedom in tasks representative of daily life.Integrating these elements, we demonstrated for the first time that an individual with an above-elbow amputation could intuitively control all five fingersof a bionic hand as if it were their own.