Comparing Analytical and AI-based Image Analysis for Micron-sized Particle Detection and Measurement

Abstract

Abstract – Microscopy is the study of analyzing cells and molecules at the microscopic level using tools such as light microscopes. Images from the microscope are imported into image analysis software to further evaluate the cell count and measurements of the cell. Artificial intelligence is being utilized in microscopy for faster analysis of cells, allowing researchers to train the software to recognize specific cells and molecules. Analytical and artificial image analysis programs, using FIJI/ImageJ and Polygon AI respectively, are compared by their ability to analyze chitosan and gold microparticles, as well as compare the speed and accuracy of the software’s identification and modeling. Results show that the Polygon AI software fares better at cell detection where the pre-trained Somatic model fares over FIJI/ImageJ yielding up to 97% true positive count for chitosan and 90% true positive count for gold microparticles. FIJI/ImageJ fares better at reporting cell measurements, where the region of interest (ROI) borders fully encapsulate the cell compared to Polygon AI with diameter measurement differences of 2.7% for chitosan and 41.9% for gold microparticles. Detection of both cells in Polygon AI takes only 10 seconds, while cell counting in FIJI/ImageJ takes up to 253 seconds. A combination of both analytical and artificial intelligence programs is needed to combine both their advantages to produce reliable results.

Keywords: microscopy, image analysis, polygon AI, FIJI-image J, cellular structures