Evaluation of the effectiveness of Gaussian filtering in distinguishing punctate synaptic signals from background noise during image analysis

• Images of synapses often contain non-specific, spatially broad background noise.
• Background can be defined by Gaussian filtering and subtracted from original image.
• This method was evaluated using hippocampal glutamatergic synapses and simulations.
• This method was efficient in background subtraction and peak detection.
• Some disadvantages were also noted, in comparison to a rolling-ball algorithm.

BackgroundImages in biomedical imaging research are often affected by non-specific background noise. This poses a serious problem when the noise overlaps with specific signals to be quantified, e.g. for their number and intensity. A simple and effective means of removing background noise is to prepare a filtered image that closely reflects background noise and to subtract it from the original unfiltered image. This approach is in common use, but its effectiveness in identifying and quantifying synaptic puncta has not been characterized in detail.New analysisWe report on our assessment of the effectiveness of isolating punctate signals from diffusely distributed background noise using one variant of this approach, “Difference of Gaussian(s) (DoG)” which is based on a Gaussian filter.ResultsWe evaluated immunocytochemically stained, cultured mouse hippocampal neurons as an example, and provided the rationale for choosing specific parameter values for individual steps in detecting glutamatergic nerve terminals. The intensity and width of the detected puncta were proportional to those obtained by manual fitting of two-dimensional Gaussian functions to the local information in the original image.Comparison with existing methodsDoG was compared with the rolling-ball method, using biological data and numerical simulations. Both methods removed background noise, but differed slightly with respect to their efficiency in discriminating neighboring peaks, as well as their susceptibility to high-frequency noise and variability in object size.ConclusionsDoG will be useful in detecting punctate signals, once its characteristics are examined quantitatively by experimenters.