Windmiller Kolster Scientific

    Software :: MRI Coils :: Consulting



  • .:. Company .:.

    Who We Are​

    Windmiller Kolster Scientific (WK+S) is a company offering technology and services to research groups engaged in brain imaging. Since our incubation at the Department of Neuroscience at the KU Leuven Medical School in 2008 with help from Leuven Research and Development (LRD) and our founding as a U.S.-based company in Fresno, CA, in 2012, WK+S has worked with a broad range of University and private clients in the US, Canada, and Europe.

    What We Do​

    We offer services in a broad range of topics related to scientific research and software engineering in medical imaging. We offer expertise in


    Software Development

    • Analysis software and data engineering including 
      • Medical image analytics (Freesurfer, SPM, FSL, Slicer).
      • EEG analysis (EEGLAB, sLORETA, MNE Python, ICA).
      • Data pipelines, numeric optimization.
      • Unit testing, version control (SVN).
    • Medical image processing including 
      • MR Image reconstruction (scanner raw data, multiband).
      • MR Image correction and alignment (EPI, fMRI).
      • Inter- and intra-modal image registration (MRI, CT, Ultrasound, Angiograms).

    MR imaging 

    • Customized MR coils and preamp systems for neuroimaging.
    • On-site consulting and training of personnel.
    • Grant support.

    We are experienced in working on a range of MR scanners including Siemens, Bruker, and Philips systems with field strengths of 3 T, 4.7 T, 7 T, and 9.4 T.


    Scientific Consulting in a variety of subjects. Past and current projects include

    • Visualizing NFL Concussions - Please check out the following webpage for our support of a project at BodyLogicMD: https://www.bodylogicmd.com/visualizing-nfl-concussions.
    • Neurofeedback as a treatment for Dyslexia - EEG analysis using EEGLab, sLORETA, and MNE Python.

    • High-k dielectrics in capacitors -  Literature review of physical modeling of gap voltages for a patent application.
    • Retinal Perfusion in Fluorescein Angiograms - Software development for angiogram image alignment and analysis of the fluorescein signal rise and fall times in arteries and veins.
    • Ultrasound and CT image alignment. - Software development for detection of the cortical surface of the bone and cross-modal alignment.

    Find us on Upwork: https://www.upwork.com/fl/haukekolster


    Hauke Kolster, Ph.D. - Neuroscientist, Physicist, Data Engineer, Entrepreneur.



    • M.Sc. in Physics - Bonn University
    • Ph.D. in Physics - LMU Munich
    • UC Berkeley - summer session classes in quantum mechanics and underground film
    • Max Planck Institute for Nuclear Physics, Heidelberg - visiting scientist
    • DESY, Hamburg - visiting scientist

    Training and Experience

    • NIKHEF Amsterdam - postdoc in physics
    • Massachusetts Institute of Technology, LNS - postdoc in physics
    • MGH/Harvard Medical School, A.A. Martinos Center - research fellow in biomedical imaging and neuroscience
    • KU Leuven Medical School, Department of Neuroscience - research assistant professor

    Entrepreneurship, Founder

    • Windmiller Kolster Scientific (WK+S), Fresno, California, USA - Principal

    Research Highlights

    Scientific research published in Kolster et al. (2009, 2010, and 2014) on the cortical organizations of the occipital and IT cortices in the human and non-human primate is featured in a popular textbook as well as several review articles. Please find details about the articles by Kolster et al. below in the research section.


    The summary of the cortical organizations described in Kolster et al. (2010) is featured in

      • the textbook Fundamentals of Human Neurophysiology by B. Kolb and I. Whishaw, 7th Edition, 2015.
      • the review article The ventral visual pathway: An expanded neural framework for the processing of object quality; DJ Kravitz et al.; Trends Cogn Sci. 2013 Jan;17(1):26-49.

      The parcellation of the MT cluster as described in Kolster et al. (2009) is included in the Caret Atlas as parcellation KMA09 and featured in the review article

      • Cortical Parcellations of the Macaque Monkey Analyzed on Surface-Based Atlases; D Van Essen et al.; Cereb Cortex. 2012 Oct; 22(10): 2227–2240.

      The summary of the cortical organizations described in Kolster et al. (2014) is featured in the review articles

      • .:. MRI Related Software .:.

        We offer three software packages that are specifically designed for image reconstruction and alignment in functional imaging of NHPs: Offline_SENSE, Align_EPI, and Align_Anatomy. These packages represent a comprehensive image reconstruction and alignment pipeline for functional data sets in either Siemens raw data or NIFTI format to fully aligned functional data sets in NIFTI format. Each package can also be used as a stand-alone solution for individual stages of the image processing. The packages are available as MATLAB GUIs on OSX, Linux, and Windows.

        EPI Image Reconstruction

        NHP imaging on a Siemens 3T TimTrio (KU Leuven). Comparison of images using online GRAPPA and offline SENSE reconstructions based on the same raw data. In accelerated imaging, body motion causes dynamic field changes, which might lead to ghosting artifacts when using GRAPPA reconstruction. SENSE reconstruction is less prone to these effects and the offline reconstructed data have significantly less ghosting artifacts as compared to the online reconstruction.

        The software is compatible with raw data acquired on Siemens 3 T Tim Trio, Prisma, and Skyra, as well as Siemens 7 T UHF.


        .:. .:. .:.

        Image Alignment Overview

        Single- and multi-session alignment can be generally split into two steps, a local functional-to-functional (U) and a global functional-to-anatomical alignment (W).

        The local alignment represents a correction of differences between similar images of the same modality that are caused by dynamic changes in experimental conditions, e.g. the subject's body motion and is performed between an EPI image and a reference EPI image. We use an in-plane distortion correction algorithm, which is designed to apply corrections along the degrees of freedom, e.g. only within an image slice and along the phase-encoding direction, and thus only reverse the distortions that are present in the EPI images.

        The global alignment represents a correction of differences between images of different modalities and is performed between the average EPI volume of a session and the anatomical MPRGAE volume and focuses on static distortions such as susceptibility gradients that are common to all images.

        In multi-session experiments, additional global alignments (V1, V2, ...) are performed between the average EPI images of the individual sessions and a reference session.


        .:. .:. .:.

        EPI Time-Course Alignment

        The parameters of a rigid-body motion correction (top) show strong translations only in phase encoding direction (green). This is expected in a model in which the translations observed in the images are created by dynamic in-plane field gradients that cause the shift of the MR image and not by actual head motion. These shifts only occur in the phase-encoding direction as the frequency-encoding and slice-select directions are not significantly affected by the dynamic gradients.

        A re-run of the rigid-body motion corrections after non-rigid image correction shows that all translations in phase encoding directions (green) are successfully corrected and that the other two dimensions are not affected.

        Temporal SNR

        Rigid vs. Non-Rigid Body Corrections

        Comparison of single image SNR (top), temporal SNR after rigid body motion correction (middle), and temporal SNR after in-plane distortion correction (bottom) for the different acceleration factors in the same subject. In non-accelerated imaging (R=1) the temporal SNR values resulting from an in-plane distortion correction are twice as high as compared to a conventional rigid motion correction (here SPM, 6 parameters). In accelerated imaging, gains in average tSNR values of 50% (R=2) and 35% (R=3) can be achieved using an in-plane motion correction instead of a rigid-body motion correction. These gains in tSNR directly translate into gains of the same magnitude in the t-scores of an fMRI analysis and gains in the t-scores of more than a factor of two have been observed in small gray matter ROI's.


        .:. .:. .:.

        The following software packages for image reconstruction and functional-to-functional and functional-to-anatomical alignment are available as MATLAB Apps. Software packages 1-3 are available for Linux and OSX. In addition, software packages 2 and 3 are available for Windows.

        Package 1: Offline EPI Reconstruction

        Offline_SENSE (MATLAB App)

        Offline SENSE reconstruction of EPI time series based on multi-channel EPI and GRE raw data acquired on a Siemens scanner. This EPI image reconstruction package features a regularized SENSE image reconstruction algorithm and an advanced image-based EPI phase correction algorithm to minimize N/2 ghosting artifacts. It further allows to either use the intrinsic EPI reference images or external GRE images as sensitivity maps for image reconstruction.

        Package 2: In-plane Distortion Correction

        Align_EPI (MATLAB App)

        Non-rigid MRI image distortion correction and alignment tool for individual images within an EPI time course. Image volumes are aligned in-plane and slice-by-slice with corrections applied only in phase encoding direction.

        Package 3: Functional to Anatomical Alignment

        Align_Anatomy (MATLAB App)

        Alignment tool for functional to anatomical alignment of EPI time course data making use of in-session GRE and structural volumes.

      • .:. Our MRI Solutions at Work .:.

        The following images demonstrate the precision that can be achieved in the mapping of functional activation to anatomical volume when using our software packages. This work was performed on a Siemens 3T Tim Trio with 1 mm isotropic resolution using the same technology as offered in our company including MR coils as well as software for image reconstructions, distortion correction, and image alignment. Figures are reproduced from Kolster et al.; JNeurosci (2014), 34(31) 10168.

        Anatomical​ Volume

        MPRAGE volume with main and pial surfaces as determined by Freesurfer.

        Alignment of Functional Volumes

        Average EPI volume with main and pial surfaces based on the MPRAGE volume. Individual EPI volumes were distortion corrected and aligned to the structural volumes using algorithms embedded in WK+S proprietary software packages.

        Functional Overlay

        Functional overlay (% signal change) with average EPI volume. The detected cortical activity coincides well with locations of grey matter as indicated by the lines representing the main and pial surfaces.


      • .:. Contact Us .:.

        Please fill out the following form and we will respond as soon as possible.

        Hauke Kolster, Ph.D.
        Fresno, CA
        (559) 312-8466
      • .:. Connect With Us .:.

      • Mavenlink Project Manager

        Existing customers with an active Mavenlink login can use the link below to login to their WK+S project.

      • .: Research :.

        The following research papers give an overview of the retinotopic organization of visual areas in the occipital lobe and the occipitoparietal junction in humans (18 areas) and the occipitotemporal junction in non-human primates (NHP) (12 areas). At the occipitotemporal junction, this organization includes the MT/V5 cluster, which is comprised of areas MT/V5, MSTv, (ph)FST, and (ph)V4t, as well as the areas V4A, OTd, PITd, and PITv in the NHP with putative homologs LO1, LO2, phPITd, and phPITv, in the human. At the occipitoparietal junction in the human, this includes the four areas of the human V3A complex, V3A and V3B as well as V3C and V3D, which form two distinct field map clusters.

        Correspondences between retinotopic areas and myelin maps in human visual cortex.

        RO Abdollahi*, H Kolster*, MF Glasser*, et al.; Neuroimage, October 1, 2014 • 99:509–524 (* co-first authors)

        Abstract: We generated probabilistic area maps and maximum probability maps (MPMs) for a set of 18 retinotopic areas previously mapped in individual subjects (Georgieva et al., 2009 and Kolster et al., 2010) using four different inter-subject registration methods. The best results were obtained using a recently developed multimodal surface matching method. The best set of MPMs had relatively smooth borders between visual areas and group average area sizes that matched the typical size in individual subjects. Comparisons between retinotopic areas and maps of estimated cortical myelin content revealed the following correspondences: (i) areas V1, V2, and V3 are heavily myelinated; (ii) the MT cluster is heavily myelinated, with a peak near the MT/pMSTv border; (iii) a dorsal myelin density peak corresponds to area V3D; (iv) the phPIT cluster is lightly myelinated; and (v) myelin density differs across the four areas of the V3A complex. Comparison of the retinotopic MPM with cytoarchitectonic areas, including those previously mapped to the fs_LR cortical surface atlas, revealed a correspondence between areas V1–3 and hOc1–3, respectively, but little correspondence beyond V3. These results indicate that architectonic and retinotopic areal boundaries are in agreement in some regions, and that retinotopy provides a finer-grained parcellation in other regions. The atlas datasets from this analysis are freely available as a resource for other studies that will benefit from retinotopic and myelin density map landmarks in human visual cortex.

        (2014) The Retinotopic Organization of Macaque Occipitotemporal Cortex Anterior to V4 and Caudoventral to the Middle Temporal (MT) Cluster.

        H Kolster et al.; The Journal of Neuroscience, July 30, 2014 • 34(31):10168 –10191

        Abstract: The retinotopic organization of macaque occipitotemporal cortex rostral to area V4 and caudorostral to the recently described middle temporal (MT) cluster of the monkey (Kolster et al., 2009) is not well established. The proposed number of areas within this region varies from one to four, underscoring the ambiguity concerning the functional organization in this region of extrastriate cortex. We used phase-encoded retinotopic functional MRI mapping methods to reveal the functional topography of this cortical domain. Polar-angle maps showed one complete hemifield representation bordering area V4 anteriorly, split into dorsal and ventral counterparts correspond- ing to the lower and upper visual field quadrants, respectively. The location of this hemifield representation corresponds to area V4A. More rostroventrally, we identified three other complete hemifield representations. Two of these correspond to the dorsal and the ventral posterior inferotemporal areas (PITd and PITv, respectively) as identified in the Felleman and Van Essen (1991) scheme. The third representation has been tentatively named dorsal occipitotemporal area (OTd). Areas V4A, PITd, PITv, and OTd share a central visual field representation, similar to the areas constituting the MT cluster. Furthermore, they vary widely in size and represent the complete contralateral visual field. Functionally, these four areas show little motion sensitivity, unlike those of the MT cluster, and two of them, OTd and PITd, displayed pronounced two-dimensional shape sensitivity. In general, these results suggest that retinotopically organized tissue extends farther into rostral occipitotemporal cortex of the monkey than generally assumed.

        The Retinotopic Organization of the Human Middle Temporal Area MT/V5 and Its Cortical Neighbors.

        H Kolster et al.; The Journal of Neuroscience, July 21, 2010 • 30(29):9801–9820

        Abstract: Although there is general agreement that the human middle temporal (MT)/V5+ complex corresponds to monkey area MT/V5 proper plus a number of neighboring motion-sensitive areas, the identification of human MT/V5 within the complex has proven difficult. Here, we have used functional magnetic resonance imaging and the retinotopic mapping technique, which has very recently disclosed the organization of the visual field maps within the monkey MT/V5 cluster. We observed a retinotopic organization in humans very similar to that documented in monkeys: an MT/V5 cluster that includes areas MT/V5, pMSTv (putative ventral part of the medial superior temporal area), pFST (putative fundus of the superior temporal area), and pV4t (putative V4 transitional zone), and neighbors a more ventral putative human posterior inferior temporal area (phPIT) cluster. The four areas in the MT/V5 cluster and the two areas in the phPIT cluster each represent the complete contralateral hemifield. The complete MT/V5 cluster comprises 70% of the motion localizer activation. Human MT/V5 is located in the region bound by lateral, anterior, and inferior occipital sulci and occupies only one-fifth of the motion complex. It shares the basic functional properties of its monkey homolog: receptive field size relative to other areas, response to moving and static stimuli, as well as sensitivity to three-dimensional structure from motion. Functional properties sharply distinguish the MT/V5 cluster from its immediate neighbors in the phPIT cluster and the LO (lateral occipital) regions. Together with similarities in retinotopic organization and topological neighborhood, the functional properties suggest that MT/V5 in human and macaque cortex are homologous.

        Visual Field Map Clusters in Macaque Extrastriate Visual Cortex.

        H Kolster et al.; The Journal of Neuroscience, May 27, 2009 • 29(21):7031–7039

        Abstract: The macaque visual cortex contains 30 different functional visual areas, yet surprisingly little is known about the underlying organizational principles that structure its components into a complete “visual” unit. A recent model of visual cortical organization in humans suggests that visual field maps are organized as clusters. Clusters minimize axonal connections between individual field maps that represent common visual percepts, with different clusters thought to carry out different functions. Experimental support for this hypothesis, however, is lacking in macaques, leaving open the question of whether it is unique to humans or a more general model for primate vision. Here we show, using high-resolution blood oxygen level-dependent functional magnetic resonance imaging data in the awake monkey at 7 T, that the middle temporal area (area MT/V5) and its neighbors are organized as a cluster with a common foveal representation and a circular eccentricity map. This novel view on the functional topography of area MT/V5 and satellites indicates that field map clusters are evolutionarily preserved and may be a fundamental organizational principle of the Old World primate visual cortex.

        In vivo and ex vivo 19-fluorine magnetic resonance imaging and spectroscopy of betacells and pancreatic islets using GLUT-2 specific contrast agents

        S. Liang, K. Louchami, H. Kolster, et al.; Contrast Media Mol. Imaging 2016, 11 506–513

        Abstract: The assessment of the β-cell mass in experimental models of diabetes and ultimately in patients is a hallmark to understand the relationship between reduced β-cell mass/function and the onset of diabetes. It has been shown before that the GLUT-2 transporter is highly expressed in both β-cells and hepatocytes and that D-mannoheptulose (DMH) has high uptake specificity for the GLUT-2 transporter. As 19-fluorine MRI has emerged as a new alternative method for MRI cell tracking because it provides potential non-invasive localization and quantification of labeled cells, the purpose of this project is to validate β-cell and pancreatic islet imaging by using fluorinated, GLUT-2 targeting mannoheptulose derivatives (19FMH) both in vivo and ex vivo. In this study, we confirmed that, similar to DMH, 19FMHs inhibit insulin secretion and increase the blood glucose level in mice temporarily (approximately two hours). We were able to assess the distribution of 19FMHs in vivo with a temporal resolution of about 20minutes, which showed a quick removal of 19FMH from the circulation (within two hours). Ex vivo MR spectroscopy confirmed a preferential uptake of 19FMH in tissue with high expression of the GLUT-2 transporter, such as liver, endocrine pancreas and kidney. No indication of further metabolism was found. In summary, 19FMHs are potentially suitable for visualizing and tracking of GLUT-2 expressed cells. However, current bottlenecks of this technique related to the quick clearance of the compound and relative low sensitivity of 19F MRI need to be overcome.