A histology-based model of quantitative T1 contrast for in-vivo - PowerPoint PPT Presentation

A histology-based model of quantitative T1 contrast for in-vivo cortical parcellation of high-resolution 7 Tesla brain MR images J. Dinse 1,2 , M. Waehnert 1 , C. L. Tardif 1 , A. Schäfer 1 , S. Geyer 1 , R. Turner 1 , P.-L. Bazin 1 Presented by Juliane Dinse 1 Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany 2 Simulation and Graphics Department, Faculty of Computer Science, Otto-von-Guericke University Magdeburg, Germany Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Cortical anatomy and cytoarchitecture Brodmann ‘ s Map, Cytoarchitecture, Atlas of von Economo Human brain 1909, lateral view Vogt, 1903 and Koskinas, 1925, (43 areas) (107 areas in total, 40 areas quantified) Cytoarchitectonic m apping of Brodm ann Areas ( BA) is accepted as standard reference. Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Myeloarchitecture Cell stain Myelin stain ? Myeloarchitecture Myeloarchitectonic Incomplete myelo- Vogt, 1903 map, E. Smith, architectonic map, 1907, lateral view Vogt & Vogt, 1910, frontal pole Research in this field is incom plete, inconclusive or even contradictory. Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Motivation  Cytoarchitecture can be transformed into information regarding relative cortical myelin density  Cortical myelin provides MRI contrast: Enables segregation of primary areas based on cortical profiles (Geyer et al., 2011; Dinse et al., 2013) Myeloarchitecture Model Cytoarchitecture Cortical depth Hellwig, 1993 2400 1400 T1 (ms) T1 m ap, 0 .5 3 m m , courtesy of M. Waehnert, 2013 Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Is there a way of mapping myeloarchitecture in-vivo onto the human cortical surface? BA 3b, layer IIIc: Step 1 Thick (%): 10 Cells: 30 Cell Size: 17 Step 2 Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

General assumptions Assum ption I : Assum ption I I : Cell size is proportional to myelin Horizontal pattern originates from concentration. axonal collaterals of cells. (Paldino & Harth, 1975) Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Step1: Generate myelin density profiles  Obtain quantitative measures of cellular configuration of each cortical layer in each ROI (von Economo & Koskinas, 1925)  Link measures to assumption I  First estimate of myelin density  Convolve graph with model given in assumption II (Paldino, 1975)  Qualitative indicator of myelin concentration in our ROIs = * * Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Step 2: Normalization to T1 contrast  Define individual range of T1 values for each ROI  Normalize profiles into T1 contrast of gray matter (Rooney et al., 2007)  Convolve with Lorentzian kernel to account for MR limiting effects (partial voluming and resolution)  Quantitative indicator of myelin concentration in our ROIs = * Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Comparison between model and data Empirical Data Modelled Profile MR adjusted Mod. Profile BA 3b BA 4 BA 1 BA 2 Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Probabilistic Model Com parison Data vs Model W eighting Fct Scaling factor frequency frequency probability probability Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Data Acquisition 1 2 3 b 4 Brodmann Areas 4, 3b, 1 and 2 in primary motor-somatosensory cortex  9 subjects scanned with a 7 Tesla scanner and MP2RAGE sequence Marques et al., 2010; Hurley et al., 2010  0.5 mm isotropic T1 map with strong intra-cortical contrast in ROIs Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Processing  Rigid image registration to MNI space at (0.4 mm) 3  MGDM whole brain segmentation  CRUISE cortical surface extraction Han et al., NeuroImage 2004; Bazin et al., NeuroImage 2013  Cortical layering and profile sampling Waehnert et al., NeuroImage 2013  Manual labelling in ROIs W M/ GM and GM/ CSF boundaries Manual labels follow ing m acro- Cortical layering anatom ical landm arks in ROI s Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Probabilities on cortical surface 4 3 b 1 2 BA 4 0 1  If model and area match, probabilities are high  Surfaces show inconsistent patterns w hen m odel and area do not m atch  More details are on my poster (Wednesday, 2 – 4.30 pm, O4-01) Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Results on subject- and group level Single subject Labelled ROIs BA 4 BA 3 b BA 1 BA 2 Modelled BAs BA 4 0.87 (0.55) 0.0 (0.02) 0.83 (0.31) 0.86 (0.36) BA 3 b 0.71 (0.77) 0.87 (0.26) 0.77 ( 0.45) 0.64 (0.55) BA 1 0.19 (0.63) 0.66 (0.94) 0.89 (0.25) 0.81 (0.42) BA 2 0.68 (0.88) 0.0 (0.01) 0.80 (0.31) 0.83 (0.45) Group average BA 4 BA 3 b BA 1 BA 2 Modelled BAs BA 4 0.75 (0.47) 0.67 (0.33) 0.69 (0.31) 0.72 (0.36) BA 3 b 0.59 (0.51) 0.88 (0.26) 0.69 ( 0.47) 0.65 (0.47) BA 1 0.43 (0.52) 0.71 (0.54) 0.73 (0.45) 0.73 (0.39) BA 2 0.62 (0.52) 0.69 (0.31) 0.73 (0.46) 0.70 (0.45) Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Summary and Conclusion  Differentiation of closely related cortical functional areas is possible in in-vivo vo at ultra-high resolution  Generative model, which can predict quant quantitative e T1 m T1 maps aps  Prospective motion correction and optimized coils may help to further increase the image quality  For robust and automatic parcellation of many cortical areas, additional information is needed: - spatial priors and regularisation, topological constraints  New insights into the relation between myeloar oarchitec ectur ure e and cyto toarc rchite tecture re Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Thanks to: Pierre-Louis Bazin Miriam Waehnert Christine Tardif Andreas Schäfer Prof. Robert Turner Stefan Geyer Enrico Reimer Katja Reimann http:/ / m ipav.cit.nih.gov/ Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Dōmo arigatō Poster: Wednesday, 2 – 4.30 pm O4-01 Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Images: courtesy of Nina Härtwich 1,2 Model vs. Resolution Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

T1 map atlas and qSM atlas 22 subjects, (0.5 mm) 3 10 subjects, (0.5 mm) 3 Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Post-mortem analysis: MRI 1 2 3b 4 6 MP2RAGE, T1 map, 200 3 μ m Superimposed Layering covering our ROIs Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Post-mortem analysis: histology 1 1 2 2 3b 4 3b 4 6 6 Myelin stain, 2.5 μ m MP2RAGE, T1 map, 200 3 μ m Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Images: courtesy of Nina Härtwich 1,2 Model vs. Histology: BA 2 Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Images: courtesy of Nina Härtwich 1,2 Model vs. Histology: res 0.05 mm Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Images: courtesy of Nina Härtwich 1,2 Model vs. Histology: res 0.5 mm Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences

Images: courtesy of Nina Härtwich 1,2 Model vs. Histology: res 1 mm Juliane Dinse Max Planck Institute for Human Cognitive and Brain Sciences