SUMMARY

Morphology analysis of cellulose nanocrystals (CNCs) using Transmission Electron Microscopy (TEM) and Atomic Force Microscopy (AFM) images is an important step in the design and optimization of the processes employed in the manufacture and utilization of CNCs. Highly polydisperse nature and irregular rod-like shapes of CNCs are challenging the morphology analysis of CNC particles. Current protocols used in the analyses of CNC particle morphology for such microscopy images are largely manual and time-consuming, and often produce inconsistent results between different researchers. This thesis aims to improve and standardize morphology analysis for AFM and TEM images of CNCs and subsequently promote CNC characterization for different applications. To achieve that, the thesis is divided into two research tasks. As the first task of this thesis, a semi-automated image analysis framework is developed to quantify the structure (particle size and grouping) of CNCs in an accurate, consistent, and time-efficient way. This framework is implemented by introducing a graphical user interface, CNC-SMART, which utilizes different automated and semi-automated image processing workflows for AFM and TEM image analysis. CNC-SMART can expeditiously process high-throughput image data using these workflows while being minimally impacted by human error and variability. The second task is the utilization of the developed SMART system over two case studies so that the SMART approach can be adapted and improved for further research. Both case studies are a part of a recent inter-laboratory comparison research conducted by ten different research groups. SMART system was tested on a very large dataset helping develop standards for CNC characterization. As the main outcome of these studies, SMART was able to assess three gaps in the current state-of-the-art in CNC particle size measurements from TEM and AFM images. These gaps were imaging issues (e.g., variation in noise and contrast of the images), image analysis issues (e.g., analyst bias in dimensional measurement), and the inability to account for other CNC groupings. With these assessments, the microstructure quantification of CNCs and other similar shaped structures would be more reliable, representative, and faster which would also improve and accelerate the future process-structure-process studies.