Sholl analysis: a quantitative comparison of semi-automated methods.

BACKGROUND

Sholl analysis remains one of the most commonly used methods to quantify neuronal dendritic complexity and is therefore a key analysis tool in neurobiology. While initially proposed when the quantification of neuronal structure was undertaken manually, the advent of software packages allowing automated analysis has resulted in the introduction of several semi and fully automated methods to quantify dendritic complexity. Unfortunately results from these methods have not in all cases been consistent. We therefore compared the results of five commonly used methods (Simple Neurite Tracer, manual, Fast Sholl, Bitmap, and Ghosh lab) using manual analysis as a ground truth.

NEW METHOD

Comparison of four semi-automated methods to the manual method using diolistically labelled mouse retinal ganglion cells.

RESULTS

We report consistency across a range of published techniques. While the majority perform well (Simple Neurite Tracer and Fast Sholl profiles have areas under the curve within 4.5% of the profile derived using the manual method), we highlight two areas in two of the methods (Bitmap and Ghosh lab methods) where errors may occur, namely undercounting (>20% relative to the manual profile) and a second peak.

COMPARISON WITH EXISTING METHODS

Our results support published validation of the Fast Sholl method.

CONCLUSIONS

Our study highlights the importance of manual calibration of automated analysis software.