Visualization

Within the user interface of Omnisphero it is possible to evaluate the quality of an automated quantification by visualizing the coordinates of identified neurons on the overview image for one or more algorithms and to zoom into regions of interest. This serves the user to estimate the quality of an algorithm for neuronal identification, without the need of a manualy evaluated reference. Additionally, it is also possible to display only coordinates of non identified or falsely identified neurons of one or two automated methods.

The coordinates of non identified neurons enables the user to identify the automated method with the highest DP-value. Additionally close up images can help to identify neuron subpopulations which are lost due to special morphological characteristics not covered by the parameters of an algorithm. This gives valuable information to improve an algorithm. The coordinates of falsely identified neurons reveal the FP-rate of an automated method. Close up images can support information on limitations of a certain algorithm but can also help to spot errors like poor staining quality and unfocused images.

Altogether the visualizations of falsely identified or non identified neurons allows the user to estimate the accuracy and precision of an automated evaluation. A quantification of DP and FP-rates is accessible via a direct comparison against a manual evaluation.

Results
Results of the neuronal identification obtained with the ‘Neuronal Tracer’ algorithm or with other software can be visualized within Omnisphero. It is possible to display all neuronal positions of an algorithm on the overview image (1a) as well as on image extracts by zooming into regions of interest (1b). Quantification of neuronal positions can be used to plot concentration response curves (1c). In case a gold standard (manual evaluation) is available also only false negatives (2a) and false positives (3a) can be displayed on the overview images and the user is able to zoom into regions of interest (2b+3b). Quantification of these neuronal positions allows to plot the DP (2c) and FP- rates (3c) of different algorithms. Plotted results are shown as mean±SEM. Significant differences among one method are indicated as * and significant differences between self-developed algorithms and the Neuronal Profiling software as #.