Computational processing of optical measurements of neuronal and synaptic activity in networks.

Fluorescent reporter

All procedures involving animals were carried out according to the UK Animals (Scientific Procedures) Act 1986 and approved by the UK Home Office. We made transgenic zebrafish expressing the synaptically localised fluorescent calcium reporter SyGCaMP2 under a promoter specific for ribbon synapses ( Dreosti et al., 2009 , and Fig. 1 a). Fish were kept at a 14:10 h light:dark cycle and bred naturally. Larvae were kept in 200 μ M 1-phenyl-2-thiourea (Sigma) from 28 h post-fertilisation on to inhibit melanin formation ( Karlsson et al., 2001 ). For imaging, larvae 7–8 days post-fertilisation (dpf) were anaesthetised with 0.016% MS 222 (Sigma) and immobilised in 2.5% low melting agarose (Biogene) on a glass coverslip.

Imaging

Retinae of transgenic fish were imaged in vivo using a custom-built 2-photon microscope ( Tsai et al., 2002 ) equipped with a mode-locked Chameleon titanium–sapphire laser tuned to 915 nm (Coherent). The objective was an Olympus LUMPlanFI 40 × water immersion (NA 0.8). Emitted fluorescence was captured by the objective and by a sub-stage oil condenser, and in both cases filtered by a HQ 535/50GFP emission filter (Chroma Technology) and a “hot mirror” that reflects wavelengths > 700 nm (Edmund Optics) before detection by PMTs (Hamamatsu). Scanning and image acquisition were controlled using ScanImage v. 3.0 software ( Pologruto et al., 2003 ) running on a PC. Light stimuli were delivered by amber and blue LEDs (Luxeon) projected through the objective onto the retina. Light stimulation was controlled through Igor Pro v. 4.01 software (WaveMetrics, Lake Oswego, OR) running on a Macintosh and time locked to image acquisition. Image sequences were acquired at 1 ms per line using 64 × 64 or 128 × 100 pixels per frame. Frames from a typical recording are shown in Fig. 2 .

User interface

The software environment we have used to implement SARFIA is Igor Pro (version 6.1, Wavemetrics) which provides extensive signal and image processing capabilities, low-level programming control together with high-level analysis and presentational functions. The programming language of Igor Pro is learnt relatively easily. This environment is popular amongst electrophysiologists, with extensions such as Neuromatic and NClamp, Patcher’s Power Tools and Slice (all of which are freely available from http://www.wavemetrics.com/users/tools.htm ). The custom control panels ( Supplementary Fig. S1 ) allow point-and-click access to all major image manipulation, analysis and graphing functions.

Data import and pre-processing

Image stacks generated by ScanImage (.tiff format) were imported into Igor Pro and stored as single-precision floating-point arrays (which are termed “waves” in Igor Pro). ScanImage stores information such as the scanning speed, zoom factor, stage position, date and time of the experiment in the header of the .tiff files, and these were stored as “notes” associated with the respective waves. Since we had empirically measured the side length of an acquired image on our setup, the correct scaling of all three dimensions was calculated and immediately applied to the respective waves. In many experiments, the images in a time-series were registered to correct for motion artefacts, which often occur during in vivo experiments ( Greenberg and Kerr, 2009; Mukamel et al., 2009 ) using Igor Pro’s inbuilt image registration operation, which is based on an algorithm described by Thevenaz et al. (1998) . Frames from a typical experiment representing raw data (after image registration) are shown in Fig. 2 .