Determination of particle size distribution by image analysis software.

• Main Author: Wong, Esther Yi Ning.
• Other Authors: Kunn Hadinoto Ong
• Format: Final Year Project
• Language: English
• Published: 2009
• Subjects: DRNTU::Engineering::Chemical engineering::Biotechnology
• Online Access: http://hdl.handle.net/10356/17126
• Institution: Nanyang Technological University

The project aims to determine the particle size distribution (PSD) of spray-dried particles using automated image analysis with ImageJ software, as well as to justify if image analysis using ImageJ is comparable to that obtained from a Particle Sizer (Malvern, Mastersizer 2000) based on laser diffraction. Validation of automated particle analysis was done against manual particle measurements in ImageJ, as a reference standard. Comparisons between statistical parameters like mean volume diameter, dv and volume median diameter, dv,50 showed the degree of accuracy. Comparison of standard deviation, σ, to further assess precision was carried out by an F-test. Number-based PSD were also plotted for each method. Results show that the Particle Sizer over-estimates particle size, resulting in significant differences when compared to image analysis. Automated image analysis works well for particles with little overlap using measurements of projected area diameter, dA, while manual measurements in ImageJ should be used to analyze heavily-agglomerated particles using measurements of Feret’s diameter, dFeret. Optimal thresholding methods for image segmentation were also found. Mixture modeling is suitable for bimodal gray-value distributions, and a local threshold is usually located at the valley or end of plateau before another peak. The problem of overlapping particles and agglomerates is also investigated to improve the accuracy of its image analysis.