Dockerized Pipeline for Analytics on Microcopy and Profilometry Data

Objectives: With the automation of microscopy and profilometry systems, both in terms of sample loaders and collection point programming, the amount of information that can be collected can be hard to analyze. With dental materials studies looking at multimodal data sets on the samples, a processing pipeline was desired that could ingest the machine files and produce the data dashboard. By using a contained code solution (Docker), the resultant program could also be deployed to high performance computing (HPC) systems to enable large scale data analysis.
Methods: 75 enamel samples were prepared for an orthodontic demineralization analysis that sought to evaluate primer product performance: 5 products with 15 samples per group. Sputter coating of the samples in a vacuum apparatus for 4 min produced a 10-50μm gold-platinum layer needed for analysis. A Quanta 3D FIB-SEM was utilized to examine the samples. General views of the samples were taken at 25-50x (5kV,WD25-35mm) and focus points were imaged at 3500x (5kV,8-12mm). Five focus points were imaged within the treatment zone on the sample: 1 center point and 4 corner points. EDS was performed on all SEM views; both general and focus points. Python was chosen as the programming environment for the solution based on its prevalence and license-free distribution. Docker was used for the deployment of the solution and LaTex was used for document generation.
Results: Dashboards were created for each of the treatment group samples that fused the SEM data, EDS data, and can accommodate profilometry data. Routines to perform statistical analysis across the study were included to encompass Kolmogorov-Smirnov, Two-Way ANOVA, and Tukey HSD (Post-Hoc). A document generator can produce the dashboards in PDF format and there is a Jupyter Notebook dynamic data visualization.
Conclusions: A Dockered code solution for large scale microscopy dataset processing and visualization enables conformity and scalability.