Methodology for Increasing Image Feature Measurement Accuracy
Michael Paul Majurski, Joe Chalfoun, Steven Lund, Peter Bajcsy, Mary C. Brady
Motivation Image features are computed in cell biology to derive quantitative information regarding cell state, differentiation, biological activity, and cell dynamics. The accuracy of any biological conclusions depends on the accuracy of the measured feature values which depends on the image quality. Image quality is often constrained by experimental considerations. The motivation for our work is the improvement of measurement accuracy for image features extracted from time- lapse fluorescent images of stem cell colonies. Due to cell sensitivity, only brief low intensity light could be used to excite fluorophores, producing images with low signal to noise ratios. Therefore, image processing is required to mitigate the effect of this experimental constraint on the acquired images in order to accurately measure the colony features of interest. Results We present a methodology for improving the accuracy of image feature measurements by combining experimentally obtained reference signal and background noise to create pseudo-real images typical of the target experiment. These images test processing pipelines in terms of maximizing feature measurement accuracy. Our application involves the selection of a single image processing spatial filter which best improves the feature measurement accuracy. Three filters and six kernel sizes were evaluated to select a 7x7 Gaussian kernel which minimized the feature residual root mean squared error. Additionally, an image processing spatial filter is selected for each image feature resulting in higher accuracy then applying the 7x7 Gaussian kernel for all image features. Conclusions This methodology enables the empirical design and validation of the image processing pipeline required to improve image feature measurement accuracy. It is applicable across experiments, imaging conditions, processing pipelines, and imaging modalities.
Computer Vision for Microscopy Image Analysis (CVMI)
, Chalfoun, J.
, Lund, S.
, Bajcsy, P.
and Brady, M.
Methodology for Increasing Image Feature Measurement Accuracy, Computer Vision for Microscopy Image Analysis (CVMI), Las Vegas, NV, US, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=920473
(Accessed December 3, 2023)