A new paper by CNBP researcher Tony Orth (lead author pictured) describes how to use large image sets to perform cell classification and imaging performance. The work combines years of hardware development on a high throughput microlens microscope together with large scale image processing.
The authors on the paper found that they could get a computer to accurately identify white blood cells types purely from a reference set of images (or dictionary), without resorting to time-consuming manual classification by trained staff.
Moreover, the authors demonstrated that because white blood cells come in a limited number of shapes and sizes, even a very poor quality noisy image of a white blood cell can be effectively enhanced by looking for similar images in the dictionary set.
This has potential applications for low-light level imaging. Working with a small amount of light is detrimental in terms of image quality but gentle on cells. The author’s dictionary-based method provides a way to partially recover image quality for dose-limited imaging.
The paper is accessible online.
Dictionary-enhanced imaging cytometry. Antony Orth, Diane Schaak and Ethan Schonbrun. Scientific Reports 7, Article number: 43148 (2017).