Examples & Applications


Live-cell imaging assays produce image stacks of cellular changes over time. SAPHIRE can be used to model the temporal changes in phenotypes of imaged cells, such as their morphologies. The example below is from the Demo that can be found in the SAPHIRE software package. Following image processing of time series stacks, the time series editing GUI (shown below) can be used to edit segmentation masks and label image frames to designate when a cell is dividing or dying, if relevant:

                                        

 

Image-based features are extracted from the cell object at each time point, and are projected onto lower dimension using Principal Component Analysis (PCA). This results in a temporal trajectory (approximately 18 hours in this example) of the cell moving through PCA shape-space:


                                     

SAPHIRE applies a hidden Markov model and Bayesian model selection (for penalizing model complexity) to this trajectory and infers underlying states that the cell moves through over time in shape-space, while directly incorporating temporal dependencies in the data. At each point in time, the cell is annotated with the most likely state it is in:


                                     

 

In this example, shape complexity is reduced into four underlying states. The states are labeled and colored for the cell at each time point, specifying the actin cytoskeleton (gray) morphological changes that the cell undergoes during the imaging experiment:


         


The SAPHIRE models with temporal state annotations have various applications, such as for facilitating interpretation and reducing complexity of noisy time series image data, as phenotypic readouts in live-cell, high-content imaging studies for drug screening, cell spreading assays, and classification of dynamic cellular responses to perturbations of interest.