Pixels as ROIs (PAR): A Less-Biased and Statistically Powerful Approach for Gleaning Functional Information from Image Stacks

Especially in the last decade or so, there have been dramatic advances in fluorescence-based imaging methods designed to measure a multitude of functions in living cells. Despite this, many of the methods used to analyze the resulting images are limited. Perhaps the most common mode of analysis is the choice of regions of interest (ROIs), followed by quantification of the signal contained therein in comparison with another "control" ROI. While this method has several advantages, such as flexibility and capitalization on the power of human visual recognition capabilities, it has the drawbacks of potential subjectivity and lack of precisely defined criteria for ROI selection. This can lead to analyses which are less precise or accurate than the data might allow for, and generally a regrettable loss of information. Herein, we explore the possibility of abandoning the use of conventional ROIs, and instead propose treating individual pixels as ROIs, such that all information can be extracted systematically with the various statistical cutoffs we discuss. As a test case for this approach, we monitored intracellular pH in cells transfected with the chloride/bicarbonate transporter slc26a3 using the ratiometric dye SNARF-5F under various conditions. We performed a parallel analysis using two different levels of stringency in conventional ROI analysis as well as the pixels-as-ROIs (PAR) approach, and found that pH differences between control and transfected cells were accentuated by ~50-100% by using the PAR approach. We therefore consider this approach worthy of adoption, especially in cases in which higher accuracy and precision are required.