To fully understand how the brain works, it is necessary to relate the brain’s function to its anatomy. Cortical anatomy is subject-specific. It is character- ized by the thickness and number of intracortical layers, which differ from one cortical area to the next. Each cortical area fulfills a certain function. With magnetic res- onance imaging (MRI) it is possible to study structure and function in-vivo within the same subject. The resolution of ultra-high field MRI at 7T allows to resolve intracortical anatomy. This opens the possibility to relate cortical function of a sub- ject to its corresponding individual structural area, which is one of the main goals of neuroimaging. To parcellate the cortex based on its intracortical structure in-vivo, firstly, im- ages have to be quantitative and homogeneous so that they can be processed fully- automatically. Moreover, the resolution has to be high enough to resolve intracortical layers. Therefore, the in-vivo MR images acquired for this work are quantitative T1 maps at 0.5 mm isotropic resolution. Secondly, computational tools are needed to analyze the cortex observer-independ- ently. The most recent tools designed for this task are presented in this thesis. They comprise the segmentation of the cortex, and the construction of a novel equi-volume coordinate system of cortical depth. The equi-volume model is not restricted to in- vivo data, but is used on ultra-high resolution post-mortem data from MRI as well. It could also be used on 3D volumes reconstructed from 2D histological stains. An equi-volume coordinate system yields firstly intracortical surfaces that follow anatomical layers all along the cortex, even within areas that are severely folded where previous models fail. MR intensities can be mapped onto these equi-volume surfaces to identify the location and size of some structural areas. Surfaces com- puted with previous coordinate systems are shown to cross into different anatomical layers, and therefore also show artefactual patterns. Secondly, with the coordinate system one can compute cortical traverses perpendicularly to the intracortical sur- faces. Sampling intensities along equi-volume traverses results in cortical profiles that reflect an anatomical layer pattern, which is specific to every structural area. It is shown that profiles constructed with previous coordinate systems of cortical depth disguise the anatomical layer pattern or even show a wrong pattern. In contrast to equi-volume profiles these profiles from previous models are not suited to analyze the cortex observer-independently, and hence can not be used for automatic delineations of cortical areas. Equi-volume profiles from four different structural areas are presented. These pro- files show area-specific shapes that are to a certain degree preserved across subjects. Finally, the profiles are used to classify primary areas observer-independently.:1 Introduction p. 1 2 Theoretical Background p. 5 2.1 Neuroanatomy of the human cerebral cortex . . . .p. 5 2.1.1 Macroscopical structure . . . . . . . . . . . .p. 5 2.1.2 Neurons: cell bodies and fibers . . . . . . . .p. 5 2.1.3 Cortical layers in cyto- and myeloarchitecture . . .p. 7 2.1.4 Microscopical structure: cortical areas and maps . .p. 11 2.2 Nuclear Magnetic Resonance . . . . . . . . . . . . . .p. 13 2.2.1 Proton spins in a static magnetic field B0 . . . . .p. 13 2.2.2 Excitation with B1 . . . . . . . . . . . . . . . . .p. 15 2.2.3 Relaxation times T1, T2 and T∗ 2 . . . . . . . . . .p. 16 2.2.4 The Bloch equations . . . . . . . . . . . . . . . . p. 17 2.3 Magnetic Resonance Imaging . . . . . . . . . . . . . .p. 20 2.3.1 Encoding of spatial location and k-space . . . . . .p. 20 2.3.2 Sequences and contrasts . . . . . . . . . . . . . . p. 22 2.3.3 Ultra-high resolution MRI . . . . . . . . . . . . . p. 24 2.3.4 Intracortical MRI: different contrasts and their sources p. 25 3 Image analysis with computed cortical laminae p. 29 3.1 Segmentation challenges of ultra-high resolution images p. 30 3.2 Reconstruction of cortical surfaces with the level set method p. 31 3.3 Myeloarchitectonic patterns on inflated hemispheres . . . . p. 33 3.4 Profiles revealing myeloarchitectonic laminar patterns . . .p. 36 3.5 Standard computational cortical layering models . . . . . . p. 38 3.6 Curvature bias of computed laminae and profiles . . . . . . p. 39 4 Materials and methods p. 41 4.1 Histology . . . . . p. 41 4.2 MR scanning . . . . p. 44 4.2.1 Ultra-high resolution post-mortem data p. 44 4.2.2 The MP2RAGE sequence . . . . . . . . p. 45 4.2.3 High-resolution in-vivo T1 maps . . . .p. 46 4.2.4 High-resolution in-vivo T∗ 2-weighted images p. 47 4.3 Image preprocessing and experiments . . . . . .p. 48 4.3.1 Fully-automatic tissue segmentation . . . . p. 48 4.3.2 Curvature Estimation . . . . . . . . . . . . p. 49 4.3.3 Preprocessing of post-mortem data . . . . . .p. 50 4.3.4 Experiments with occipital pole post-mortem data .p. 51 4.3.5 Preprocessing of in-vivo data . . . . . . . . . . p. 52 4.3.6 Evaluation experiments on in-vivo data . . . . . .p. 56 4.3.7 Application experiments on in-vivo data . . . . . p. 56 4.3.8 Software . . . . . . . . . . . . . . . . . . . . .p. 58 5 Computational cortical layering models p. 59 5.1 Implementation of standard models . .p. 60 5.1.1 The Laplace model . . . . . . . . .p. 60 5.1.2 The level set method . . . . . . . p. 61 5.1.3 The equidistant model . . . . . . .p. 62 5.2 The novel anatomically motivated equi-volume model p. 63 5.2.1 Bok’s equi-volume principle . . . . . .p. 63 5.2.2 Computational equi-volume layering . . p. 66 6 Validation of the novel equi-volume model p. 73 6.1 The equi-volume model versus previous models on post-mortem samples p. 73 6.1.1 Comparing computed surfaces and anatomical layers . . . . . . . . p. 73 6.1.2 Cortical profiles reflecting an anatomical layer . . . . . . . . .p. 79 6.2 The equi-volume model versus previous models on in-vivo data . . . .p. 82 6.2.1 Comparing computed surfaces and anatomical layers . . . . . . . . p. 82 6.2.2 Cortical profiles reflecting an anatomical layer . . . . . . . . .p. 85 6.3 Dependence of computed surfaces on cortical curvature . . . . .p. 87 6.3.1 Within a structural area . . . . . . . . . . . . . . . . . . p. 87 6.3.2 Artifactual patterns on inflated surfaces . . . . . . . . . .p. 87 7 Applying the equi-volume model: Analyzing cortical architecture in-vivo in different structural areas p. 91 7.1 Impact of resolution on cortical profiles . . . . . . . . . . . . . p. 91 7.2 Intersubject variability of cortical profiles . . . . . . . . . . . p. 94 7.3 Myeloarchitectonic patterns on inflated hemispheres . . . . . . .p. 95 7.3.1 Comparison of patterns with inflated labels . . . . . . . . . .p. 97 7.3.2 Patterns at different cortical depths . . . . . . . . . . . . .p. 97 7.4 Fully-automatic primary-area classification using cortical profiles p. 99 8 Discussion p. 105 8.1 The novel equi-volume model . . . . . . . . . . . . . . . . . . . . .p. 105 8.2 Analyzing cortical myeloarchitecture in-vivo with T1 maps . . . . . .p. 109 9 Conclusion and outlook p. 113 Bibliography p. 117 List of Figures p. 127

The advent of unoccupied aerial systems (UAS) as disruptive technology has a lasting impact on remote sensing, geophysics and most geosciences. Small, lightweight, and low-cost UAS enable researchers and surveyors to acquire earth observation data in higher spatial and spectral resolution as compared to airborne and satellite data. UAS-based applications range from rapid topographic mapping using photogrammetric techniques to hyperspectral and geophysical measurements of surface and subsurface geology. UAS surveys contribute to identifying metal deposits, monitoring of mine sites and can reveal arising environmental issues associated with mining. Further, affordable UAS technology will boost exploration data availability and expertise in the global south. This thesis investigates the application of UAS-based multi-sensor data for mineral exploration, in particular the integration of hyperspectral imagers, magnetometers and digital cameras (covering the visible red, green, blue light spectrum). UAS-based research is maturing, however the aforementioned methods are not unified effectively. RGB-based photogrammetry is used to investigate topography and surface texture. Image spectrometers measure mineral-specific surface signatures. Magnetometers detect geomagnetic field changes caused by magnetic minerals at surface and depth. The integration of such UAS sensor-based methods in this thesis augments exploration potential with non-invasive, high-resolution, safe, rapid and practical survey methods. UAS-based surveying acquired, processed and integrated data from three distinct test sites. The sites are located in Finland (Fe-Ti-V at Otanmäki; apatite at Siilinjärvi) and Greenland (Ni-Cu-PGE at Qullissat, Disko Island) and were chosen as geologically diverse areas in subarctic to arctic environments. Restricted accessibility, unfavourable atmospheric conditions, dark rocks, debris and vegetation cover and low solar illumination were common features. While the topography in Finland was moderately flat, a steep landscape challenged the Greenland field work. These restraints meant that acquisitions varied from site to site and how data was integrated and interpreted is dependent on the commodity of interest. Iron-based spectral absorption and magnetic mineral response were detected using hyperspectral and magnetic surveying in Otanmäki. Multi-sensor-based image feature detection and classification combined with magnetic forward modelling enabled seamless geologic mapping in Siilinjärvi. Detailed magnetic inversion and multispectral photogrammetry led to the construction of a comprehensive 3D model of magmatic exploration targets in Greenland. Ground truth at different intensity was employed to verify UAS-based data interpretations during all case studies. Laboratory analysis was applied when deemed necessary to acquire geologic-mineralogic validation (e.g., X-ray diffraction and optical microscopy for mineral identification to establish lithologic domains, magnetic susceptibility measurements for subsurface modelling), for example for trace amounts of magnetite in carbonatite (Siilinjärvi) and native iron occurrence in basalt (Qullissat). Technical achievements were the integration of a multicopter-based prototype fluxgate-magnetometer data from different survey altitudes with ground truth, and a feasibility study with a high-speed multispectral image system for fixed-wing UAS. The employed case studies transfer the experiences made towards general recommendations for UAS application-based multi-sensor integration. This thesis highlights the feasibility of UAS-based surveying at target scale (1–50 km2) and solidifies versatile survey approaches for multi-sensor integration. / Ziel dieser Arbeit war es, das Potenzial einer Drohnen-basierten Mineralexploration mit Multisensor-Datenintegration unter Verwendung optisch-spektroskopischer und magnetischer Methoden zu untersuchen, um u. a. übertragbare Arbeitsabläufe zu erstellen. Die untersuchte Literatur legt nahe, dass Drohnen-basierte Bildspektroskopie und magnetische Sensoren ein ausgereiftes technologisches Niveau erreichen und erhebliches Potenzial für die Anwendungsentwicklung bieten, aber es noch keine ausreichende Synergie von hyperspektralen und magnetischen Methoden gibt. Diese Arbeit umfasste drei Fallstudien, bei denen die Drohnengestützte Vermessung von geologischen Zielen in subarktischen bis arktischen Regionen angewendet wurde. Eine Kombination von Drohnen-Technologie mit RGB, Multi- und Hyperspektralkameras und Magnetometern ist vorteilhaft und schuf die Grundlage für eine integrierte Modellierung in den Fallstudien. Die Untersuchungen wurden in einem Gelände mit flacher und zerklüfteter Topografie, verdeckten Zielen und unter oft schlechten Lichtverhältnissen durchgeführt. Unter diesen Bedingungen war es das Ziel, die Anwendbarkeit von Drohnen-basierten Multisensordaten in verschiedenen Explorationsumgebungen zu bewerten. Hochauflösende Oberflächenbilder und Untergrundinformationen aus der Magnetik wurden fusioniert und gemeinsam interpretiert, dabei war eine selektive Gesteinsprobennahme und Analyse ein wesentlicher Bestandteil dieser Arbeit und für die Validierung notwendig. Für eine Eisenerzlagerstätte wurde eine einfache Ressourcenschätzung durchgeführt, indem Magnetik, bildspektroskopisch-basierte Indizes und 2D-Strukturinterpretation integriert wurden. Fotogrammetrische 3D-Modellierung, magnetisches forward-modelling und hyperspektrale Klassifizierungen wurden für eine Karbonatit-Intrusion angewendet, um einen kompletten Explorationsabschnitt zu erfassen. Eine Vektorinversion von magnetischen Daten von Disko Island, Grönland, wurden genutzt, um großräumige 3D-Modelle von undifferenzierten Erdrutschblöcken zu erstellen, sowie diese zu identifizieren und zu vermessen. Die integrierte spektrale und magnetische Kartierung in komplexen Gebieten verbesserte die Erkennungsrate und räumliche Auflösung von Erkundungszielen und reduzierte Zeit, Aufwand und benötigtes Probenmaterial für eine komplexe Interpretation. Der Prototyp einer Multispektralkamera, gebaut für eine Starrflügler-Drohne für die schnelle Vermessung, wurde entwickelt, erfolgreich getestet und zum Teil ausgewertet. Die vorgelegte Arbeit zeigt die Vorteile und Potenziale von Multisensor-Drohnen als praktisches, leichtes, sicheres, schnelles und komfortabel einsetzbares geowissenschaftliches Werkzeug, um digitale Modelle für präzise Rohstofferkundung und geologische Kartierung zu erstellen.