Quantitative analysis of nanostructures’ shape and distribution in micrographs using image analysis

A set of mathematical descriptors was devised to describe structural relationships between nanostructures in micrographs. The ability of these descriptors to measure size, shape and distribution of nanostructures was assessed on Confocal Laser Scanning Microscope (CLSM) micrographs of treated milk, in which fat was stained. Comparisons of scores obtained from automatic descriptors and visual observation showed that these methods generate similar groupings among milk treatments. The discriminative power of these descriptors was also proved by their ability to infer group classifications on testing images. In a two-group classification, success rates were 89% and 91.9%, whereas rejection ratios were lower than 7% for a probability threshold up to 75%.

► A set of descriptors to measure distribution and size of nanostructures in micrographs.
► Descriptors proved effective in mimicking human discriminative power on milk micrographs.
► Automatic discrimination of unseen images was successful in about 90% of test samples.
► Useful insights can be gained from quantitative measurements.