The Gold Standard Paradox in Digital Image Analysis: Manual Versus Automated Scoring as Ground Truth.

CONTEXT: - Novel therapeutics often target complex cellular mechanisms. Increasingly, quantitative methods like digital tissue image analysis (tIA) are required to evaluate correspondingly complex biomarkers to elucidate subtle phenotypes that can inform treatment decisions with these targeted therapies. These tIA systems need a gold standard, or reference method, to establish analytical validity. Conventional, subjective histopathologic scores assigned by an experienced pathologist are the gold standard in anatomic pathology and are an attractive reference method. The pathologist's score can establish the ground truth to assess a tIA solution's analytical performance. The paradox of this validation strategy, however, is that tIA is often used to assist pathologists to score complex biomarkers because it is more objective and reproducible than manual evaluation alone by overcoming known biases in a human's visual evaluation of tissue, and because it can generate endpoints that cannot be generated by a human observer. OBJECTIVE: - To discuss common visual and cognitive traps known in traditional pathology-based scoring paradigms that may impact characterization of tIA-assisted scoring accuracy, sensitivity, and specificity. DATA SOURCES: - This manuscript reviews the current literature from the past decades available for traditional subjective pathology scoring paradigms and known cognitive and visual traps relevant to these scoring paradigms. CONCLUSIONS: - Awareness of the gold standard paradox is necessary when using traditional pathologist scores to analytically validate a tIA tool because image analysis is used specifically to overcome known sources of bias in visual assessment of tissue sections.

Artificial selection on egg size perturbs early pattern formation in Drosophila melanogaster.

Pattern formation in Drosophila embryogenesis has been widely investigated as a developmental and evolutionary model of robustness. To ask whether genetic variation for pattern formation is suppressed in this system, artificial selection for divergent egg size was used to challenge the scaling of even-skipped (eve) pattern formation in mitotic cycle 14 (stage 5) embryos of Drosophila melanogaster. Three-dimensional confocal imaging revealed shifts in the allometry of eve pair-rule stripes along both anterior­posterior (A­P) and dorsoventral (D­V) axes as a correlated response to egg size selection, indicating the availability of genetic variation for this buffered trait. Environmental perturbation was not required for the manifestation of this variation. The number of nuclei at the cellular blastoderm stage also changed in response to selection, with large-egg selected lines having more than 1000 additional nuclei relative to small-egg lines. This increase in nuclear number in larger eggs does not scale with egg size, however, as nuclear density is inversely correlated with egg length. Nuclear density varies along the A­P axis but does not correlate with the shift in eve stripe allometry between the selection treatments. Despite its macroevolutionary conservation, both eve stripe patterning and blastoderm cell number vary genetically both within and between closely related species.

Constrained and dimensionality-independent path openings.

Path openings and closings are morphological operations with flexible line segments as structuring elements. These line segments have the ability to adapt to local image structures, and can be used to detect lines that are not perfectly straight. They also are a convenient and efficient alternative to straight line segments as structuring elements when the exact orientation of lines in the image is not known. These path operations are defined by an adjacency relation, which typically allows for lines that are approximately horizontal, vertical or diagonal. However, because this definition allows zig-zag lines, diagonal paths can be much shorter than the corresponding horizontal or vertical paths. This undoubtedly causes problems when attempting to use path operations for length measurements. This paper 1) introduces a dimensionality-independent implementation of the path opening and closing algorithm by Appleton and Talbot, 2) proposes a constraint on the path operations to improve their ability to perform length measurements, and 3) shows how to use path openings and closings in a granulometry to obtain the length distribution of elongated structures directly from a gray-value image, without a need for binarizing the image and identifying individual objects.

Coupling visualization and data analysis for knowledge discovery from multi-dimensional scientific data.

Knowledge discovery from large and complex scientific data is a challenging task. With the ability to measure and simulate more processes at increasingly finer spatial and temporal scales, the growing number of data dimensions and data objects presents tremendous challenges for effective data analysis and data exploration methods and tools. The combination and close integration of methods from scientific visualization, information visualization, automated data analysis, and other enabling technologies -such as efficient data management- supports knowledge discovery from multi-dimensional scientific data. This paper surveys two distinct applications in developmental biology and accelerator physics, illustrating the effectiveness of the described approach.

A quantitative spatiotemporal atlas of gene expression in the Drosophila blastoderm.

To fully understand animal transcription networks, it is essential to accurately measure the spatial and temporal expression patterns of transcription factors and their targets. We describe a registration technique that takes image-based data from hundreds of Drosophila blastoderm embryos, each costained for a reference gene and one of a set of genes of interest, and builds a model VirtualEmbryo. This model captures in a common framework the average expression patterns for many genes in spite of significant variation in morphology and expression between individual embryos. We establish the method's accuracy by showing that relationships between a pair of genes' expression inferred from the model are nearly identical to those measured in embryos costained for the pair. We present a VirtualEmbryo containing data for 95 genes at six time cohorts. We show that known gene-regulatory interactions can be automatically recovered from this data set and predict hundreds of new interactions.