Feed

Digitized imaging of the spatial distribution of a targeted metal over a given area converts a continuous spectrum of data into a set of discrete pixels of intensities that ideally correlate strongly with the areal density map. This should be independent of the metal imaging modality, allowing semiquantitative analysis. In practice, correlation strength may vary between data sets obtained by different modalities. One enduring problem for all modalities is selecting a cutoff threshold intensity appropriately distinguishing the true signal from background noise. This may entail subjective choices between "conservative" thresholds prioritizing specificity over sensitivity (i.e., true positives despite possible false negatives) and "discovery-driven" thresholds prioritizing identification of putative effects for subsequent validation (i.e., true negatives despite possible false positives). Computerized data processing may help make this more objective by comparing outcomes for the set of all possible threshold values, termed cumulative threshold analysis. We address pitfalls in performing valid cumulative threshold analysis using ImageJ/Fiji for relative quantification of brain iron in mice with normal or genetically elevated iron, assessed by classical histochemistry or synchrotron X-ray fluorescence. In addition to pitfalls in choosing settings for generating analysis histograms, these include data loss with conversion between bit depths and selecting both appropriate X-axis directionality and image display range minima and maxima for analyzing multiple images. If these factors are handled appropriately, cumulative thresholding provides a powerful approach for more objective analysis of biometal imaging. This has important applications for metal imaging in research and clinical settings.