Data for Publication Exploring key ionic interactions for magnesium degradation in simulated body fluid - a data-driven approach

Abstract

This is the readme file for all data used within the publication "Exploring key ionic interactions for magnesium degradation in simulated body fluid - a data-driven approach"

Created: 25th January 2020 by Dr. Berit Zeller-Plumhoff Contact: berit.zeller-plumhoff@hzg.de

The publication is based on a number of experimental and computational data sets. These are: - microCT imaging data (raw and processed) before and after degradation - Fiji/ImageJ scripts for automated processing and read-out of the imaging data - SEM+EDX imaging data (raw and processed) - Matlab scripts for processing of the EDX data - Hydra/Medusa output files for precipitation prediction - Jupyter Notebooks for the final analysis and plotting of the processed data

Due to the large amount of imaging data (~70GB per microCT dataset, raw+processed), we are publishing only the processed data of two exemplary data sets - one prior to degradation and one after degradation of the sample. All other data is stored and will be provided upon request. Where a differentiation of datasets into 'C' and 'D' is denoted, this corresponds to the ph-adjusted and non-adjusted cases, respectively, as described in the publication.

Please find below a short description of all datasets that are made available. For better handling the data was divided into nine .zip files. The folder "microCT_sol1_sample1_degraded" contains the exemplary microCT data for the first sample degraded in the pH-adjusted solution 1, following data processing after reconstruction. All image files are in .tif format, with a 1.6 micrometer isotropic voxel size. The folder "cropped" contains the manually cropped filtered images of the 0.7 mm ROI (from the sample centre) after "script filtering.ijm". Using the trainable WEKA segmentation with the "classifier.model" file images were segmented and saved in "seg_weka" by utilising the script "Segmenting_with_weka.ijm". Some files had to be manually corrected - these were saved in "seg_weka_edited". Finally, the images were aligned along their longitudinal axis using the script "BW_Conversion_Alignment.ijm" and saved in the folder "aligned". Folders "microCT_sol1_sample1_degraded_raw" and "microCT_sol1_sample1_degraded_reco" are the raw projection data and tomogaphic reconstruction of this dataset. The folder structure for the initial scan of sample 1 (sample_C1_initial) is simimilar; however, it contains only a general segmented folder "segmented_all" with the filtered, segmented and aligned images and their respective outline (subfolder "outline") for the calculation of the initial sample surface area. The image processing was performed using the "filtering.ijm" script. Again, folders "sample_C1_initial_raw" and "sample_C1_initial_reco" are the raw projection data and tomogaphic reconstruction of this dataset. The Fiji scripts are saved in the folder "Fiji_Scripts", divided into those used for before and after degradation scans. Note that all file paths require changing for the scripts to run.The scripts "Data_Gathering_Initialscan_BZP_C_D.ijm" and was used to gather outline data from the entire sample, outline data from the ROI, and volume data from the ROI.

The file "edx_analysis.m" is the Matlab file that was used to analyse all EDX data for further processing. The variable f was saved as "EDX_results.mat". The paths are given relative to the .m file and need to be adjusted if this is moved. The file accesses the "EDX_data" folder. This folder contains the "SEMEDX" folder in which the elemental wt% maps from EDX measurements for each sample are stored as .txt files. The folder "Segmented" then contains the respective segmented residual magnesium images and the segmented degradation layer images that the Matlab file requires for its computation.

The folder "Data_Mg_Ion contains" all processed data that is then read and further processed by the respective Jupyter notebooks. The subfolders "Initial_data" and "Final_data" contain the histograms as calculated by Fiji/ImageJ for the volume (V) and area (A) assessment of the initial and degraded wires. The data structure is: # Final V data: /Final_data/histo_1.1C # Initial V data: /Initial_data/initial_histo_1.1C # Initial A data (whole sample, not 0.7mm ROI): /Initial_data/initial_outline_1.1C # Initial A data (0.7mm ROI): /Initial_data/initial_outline_ROI_1.1C The subfolder "Hydra_Medusa" contains the Mg and Ca precipitates as computed by Hydra Medusa for each solution in .csv format. The files "EperimentalData.csv" and "ph_per_sample.csv" contain the pH and temperature measurements for each sample/solution in averaged or full format. The file "EDX_results.mat" contains the mean elemental wt% for each element and its distribution along the degradation layer as computed in Matlab. The file "ion_concentrations.txt" contains the theoretical initial concentration for all ions. The precipitation-dependent corrections are given in "corr_conc.pkl" (which is calculated in the file "precipitation_graphs_publication.ipynb"). "statistics_lincorr_publication.ipynb" - this Jupyter notebook contains the results for the Student's t-test and the linear correlation of parameters "precipitation_graphs_publication.ipynb" - this Jupyter notebook creates the precipitation graphs used in the publication "graphs_publication.ipynb" - this Jupyter notebook creates all other graphs "tree_regression_publication.ipynb" - this Jupyter notebook contains the tree regression analysis using different input parameter sets and different regression models All figures generated by the Jupyter notebooks are saved in the "figures" subfolder.

Please note that for all Jupyter notebooks to work you may need to install missing modules.