DALAN : Disconnectivity Atlas for Lesion Analysis of Networks April 2018 version INTRODUCTION DALAN is a HARDI diffusion streamline atlas of 42 older (40-84 y.o., mean 61, std. dev. 11) healthy subjects that is intended to be used with lesion masks from clinical or other anatomical images to provide projected medium to long distance connectivity deficits implied by WM lesion damage. The atlas is associated with the popular Harvard-Oxford (H-O) cortical and subcortical atlas so that WM connectivity deficits can be analyzed in conjunction with corresponding GM deficits as well. The DALAN atlas is provided in both Octave/MatLab formats as well as HDF5 format for use with many other languages. Included are Octave/MatLab functions to demonstrate the use of the database as well as several NIfTI files demonstrating database connectivity coverage. We provide two versions of the atlas. The first uses the H-O atlas's 48 cortical and 8 subcortical parcels (25% incidence) in each hemisphere as starting and ending points for the HARDI streamlines in each atlas subject, where the streamlines remained inside each subject's own segmented white matter. For each pair of H-O parcels, DALAN provides a 100 x N sized table of location indices (of the 2mm H-O atlas provided) which corresponds to the path of N streamlines between those two parcel (frequently N=0). The second atlas is like the first except that we add in midline corpus callosum parcels from the Hofer-Frahm parcellation as targets for streamline. The second atlas version accounts for the fact that even with HARDI quality diffusion data, the accuracy of long distance, transcallosal streamlines to contalateral GM targets is not always good - thus this atlas allows a researcher to focus on the lesion's ipsilateral hemisphere while still accounting for overall likely lobar level destinations for transcallosal WM streamlines. ATLAS DETAILS In the DalanAtlas directory, you have a choice of 3 data formats in 6 separate zipped files (you should only unzip one atlas type ("First" or "Second") at a time). 1) Matlab version 6 .mat files : these files can be read by MatLab versions 5 to 8 and Octave versions 2 to 4. 2) Matlab version 7 .mat files: these files can be read by MatLab version 7 to 8 and Octave versions 3.2 to 4.x - these are compressed compared to version 6 .mat files and take up much less disk space while still being fast to read in. 3) RDF5 format .h5 files : these files can be read by MatLab versions 7 to 8 and most Octave versions 3 to 4. They can also be read by many other programming languages (e.g. C++, Java, Python) using standard open-source libraries. These are generally the slowest to read in but contain the most metadata. /DalanAtlas/ Directory File Names a) HO112_WMStreamlinesHOXX.* - for the the first atlas type (cortical and subcortical streamline targets) for subject # XX. b) HO112_WMStreamlinesCCXX.* - for the the second atlas type (including midline corpus callosum targets) for subject # XX. c) HO112_Cor+Sub+Cer_2mm_Fixed.nii - the target GM atlas for the first Dalan atlas. Created by merging the H-O (FSL version) cortical and subcortical 25% incidence atlases (plus the FSL cerebellum 25% atlas). d) HO112_Cor+Sub+Cer_2mm_Fixed_HF.nii - the target GM atlas for the second Dalan atlas. Created by adding in to the first target GM atlas the Hofer Frahm midline corpus callosum parcels (measured using C8: nitrc.org/project/C8C8) for 3 voxels straddling the midline. e) ho112_*.txt - H-O code abbreviations for the above GM target files. Variables included in each subjects HO116_WMStreamlines*.* atlas file a) H - 1x12 vector of basic properties of the target GM file [#voxels voxelsizes datatype imagecenter] b) wm - white matter codes in the target GM files c) wmidx - NIfTI indices of the target GM parcel files for the WM locations occupied by some streamline. d) wmxyz - indices of x y z locations of wmidx indices e) WMPath - a cell matrix where each WMPath{p,q} contain the streamlines connecting GM parcel #p to GM #q, such that each WMPath{p,q} is a 100xN matrix of WM indices with N being the number of streamlines connecting the parcels. this variable is in the .mat files only f) Sz - a matrix of the number of streamlines connecting the corresponding GM parcels in the target GM atlas. this variable is in the .h5 files only g) WMPath_xxx_yyy - A 100xN matrix WM indices with N being the number of streamlines connecting parcel #xxx to parcel #yyy. this variable is in the .h5 files only WHITE MATTER IMPORTANCE MAPS In the ImportanceMap directory we have provided NIfTI files showing at each likely WM voxel the percentage of connectivity reduced to/from a lobar-size area (specified by the file name) according to the DALAN streamline atlas if that voxel alone is lesioned. Of course, given that one streamline can be lesioned in multiple locations, one cannot just overlay lesion maks on importance maps and sum the values to figure out the disconnectivity, but they still provide a view of streamline coverage and the relative importance of locations vis-a-vis targeted GM areas. Files in the ImportanceMaps directory are either of the LH or RH hemisphere, and single streamline endpoint targeted to the Basal Ganglia, Brainstem, Lateral Frontal, Medial Frontal, Insula, Limbic system, Occipital, Lateral Parietal, Medial Parietal, Temporal, the Thalamaus, or the whole hemipshere. PROVIDED OCTAVE/MATLAB FUNCTIONS We provide some Octave/Matlab functions to demonstrate data extraction and use of the streamline atlases. a) test.m - this gives a demonstration of analyzing a sample of 23 real, anonymized lesion masks with the atlas (in the Lesions directory). You can use this to estimate the time it takes to use the atlas to estimate connectivity deficits (generally a few minutes per lesion mask). b) extractDalanAtlas.m - demonstrates how to extract DALAN atlas data from the .mat or .h5 subject files. c) analyzeLesions.m - uses the DALAN atlas to analyze lesion mask NIfTI files to produce basic WM estimated connectivity and GM estimated parcel deficits. d) meanwNaN.m; sumwNaN.m; padZeros.m - auxilliary functions e) NIFTI directory - Jimmy Shen's matlab toolbox for NIFTI file reading and writing LICENSE Attribution - it would be swell if you reference the nitrc.org/projects/dalan webpage if you use our atlas. You could also reference our SfN poster that compares using DALAN in stroke patients to having HARDI data for those same stroke patients (doi: 10.7490/f1000research.1113523.1). Timothy Herron US Veterans Affairs Martinez Clinic tjherron at ebire dot org All opinions contained herein are those of the authors and not those of the US Department of Veterans Affairs nor of the United States Government.