The superior colliculus (SC) is a paired and layered structure located on the dorsal surface of the midbrain playing an important role in auditory, visual, motor control, and multisensory integration. The small size of the nuclei, the composition of the midbrain, and the proximity of the SC to the brain vascular and ventricular formations are major hindrances in the way of studying human SC in-vivo. However, recent advances in magnetic resonance imaging (MRI) in terms of resolution, acquisition time, and signal-to-noise ratio (SNR) which are the products of the improved sequences and stronger magnets e.g. Ultra-High-Field (UHF) MRI has enabled the possibility of studying brain areas such as SC as a region of interest (ROI). In the first phase of this project, we estimated the global functional connectivity network (FCN) of the SC. To this end, we used 3T functional MRI (fMRI) data acquired by the Human Connectome Project (HCP) consortium. The data was preprocessed by the HCP pipeline and was corrected for motion artifacts and was registered to a standard space template. Driven by the limitation imposed by the SNR and the resolution, we studied all the SC layers of each hemisphere combined. Our results were following major animal studies while introducing unnoticed connections to frontal attentional and motor control areas. Importantly and in general, we mapped the entire functional network of the alive human brain SC, in-vivo; provide the chance for a direct comparison with previously detected structural connections of the SC in mammalians' brains. In the second phase of the project, we examined a layer-wise connectivity pattern of the human SC function in-vivo, using the 7T fMRI dataset from the HCP database. Our goal in the second phase was to identify the functional sub-networks that the SC is involved in by breaking the previously estimated network into smaller regions groups based on depth-dependent SC FCN. By spatial upsampling of the SC region and the use of k-means clustering, we discovered six sub-networks with unique characteristics. These networks comprised areas associated with vision, the oculomotor system with an emphasis on the peripheral vision, and planning goal-directed actions. Also, we identified clusters that are representing the areas of no functional connection to SC. In the third phase of this project, we developed a solution for human midbrain segmentation with a particular focus on SC. Creating an in-vivo map of the human midbrain nuclei in a standard space is essential to encourage further functional and structural studies of the regions within this area. To this end, we used a UHF MRI dataset of ten subjects including R2* and Quantitative Susceptibility Mapping (QSM), preprocessed, resliced, and standardized spatially. By fusing the information of two modalities and ten subjects and using manually selected image features we successfully mapped major midbrain nuclei including SC. Calculated features were fed into a deep learning network to create the compressed feature representation of the data and was used in hybrid with a k-means clustering machine to improve the clustering assignment confidence simultaneously with the relevance of the deep features representation. We identified the borders of SC, pars reticulata, pars compacta, and pars lateralis of the substantia nigra, magnocellular and parvocellular red nucleus, corticospinal, corticopontine and fronto-pontine crus cerebri tracts, periaqueductal grey, and inferior colliculus (IC). Our results reveal the widespread pattern of the human SC neural correlates in the resting state and the relative transformation of the function with respect to the SC depth orthogonal to the tectal plate. Data-driven segmentation of the SC connectivity map splits it up into sub-networks, each comprising areas mainly known for their involvement in peripheral vision, oculomotor system, goal-directed hand actions, and planning. We found no evidence of the human SC being functionally connected to the focal primary visual area in the resting-state. Furthermore, using the UHF structural MRI data we could delineate the borders of the human SC together with the borders of the SN pars reticulata, pars compacta, and pars lateralis, magnocellular and parvocellular areas of the RN, corticospinal, corticopontine, and fronto-pontine crus cerebri tracts, PAG and IC. Keywords: superior colliculus, midbrain, depth-dependent, fMRI, clustering, deep learning, quantitative structural mapping. Dissertation ist gesperrt bis zum 20.07.2023 !