The main tool to load the LabView data is CalciumAnalysis.ipynb
- Double check and adjust the settings in the Settings cell.
- Run the whole script by pressing Run above. Or run cell-by-cell by pressing CTRL+ENTER.
Starting the script will open a file dialog (possibly hidden behind other windows). Select the Labview .lvm file to analyze. (Only tested for "CalciumRecordings2channels.vi", 2 signal channels + stimulation channel.) - Check for error messages below each cell, and make necessary changes.
Typical data issues:- Problem: detected stimulus frequency/block lenght/ ON/OFF time is wrong.
- Cause: Beginning or end of recording contains incomplete stim block that messes up the regular paradigm.
- Solution: Set CutStart or CutEnd accordingly to remove any incomplete stim. blocks.
- Cause: Beginning or end of recording contains incomplete stim block that messes up the regular paradigm.
- Problem: detected stimulus frequency/block lenght/ ON/OFF time is wrong.
- Problem: Detected modulation frequency is wrong (e.g. both channels show 980Hz)
- Cause: There was too much laser crosstalk.
- Solution: Manually set a value for mod_ch1 or mod_ch2.
- Cause: There was too much laser crosstalk.
- Problem: Not all or too many signal drops / jumps are detected.
- Cause: Unusually high/low noise in the signal.
- Solution: adjust the advanced parameters xSTD, i.e. how many standard deviations are considered to be a jump. and timetoconsider i.e. how many seconds before and after each time point should be compared. Tip: The output of the cell Fixing jumps in data shows the time in seconds where each jump is detected. Compare by eye whether these time points make sense. Also it may need a lot of trial and error, better go directly to the Fixing jumps cell below, and play around with different values and re-run only that cell.
- Cause: Unusually high/low noise in the signal.
- Problem: Detrending is overfitting the time course or not removing enough baseline oscillations.
- Cause: For long stimulation blocks or lots of physiological noise the automaticpolynomial degree may be wrong.
- Solution: Go to the Detrending / Baseline drift removal cell, check the output *polynomial degree x. Locate the line #pnum=n in the code above, uncomment the line and enter a value larger or smaller than in the output (must be some integer value).
- Solution: Go to the Detrending / Baseline drift removal cell, check the output *polynomial degree x. Locate the line #pnum=n in the code above, uncomment the line and enter a value larger or smaller than in the output (must be some integer value).
- Cause: For long stimulation blocks or lots of physiological noise the automaticpolynomial degree may be wrong.
- To to clear all previous values and prevent errors, run whole script again after changing a parameter (Button with two arrows above, confirm when asked to restart Kernel)
- Outputs (created in the same folder as the .lvm file):
- A .mat file with the same filename as the .lvm file selected. See Matlab cell below for description of its contents.
- Plots are not stored. Manually format them using plotly and store as image.
- Two text files, regressor1_married.1D and regressor2_married.1D are created. Can be used as inputs for AFNI 3DDeconvolve. (Currently fixed at 1s resolution, like usual fMRI TR, and starting from (first stimulus event - bsl) selected in the Settings)
Limitations:
- Although it should work for any modulation frequency, the filter settings have only been tested for 1230Hz (channel1) and 980Hz (channel2).
- The automatic stimulus detection and block averaging only works for regular block stimulation, i.e. constant stim.freq., block duration, and inter-stimulus interval during the whole scan.
- Currently its not possible to cut out time periods in the middle of the time course (only start or end), but bad blocks can be removed from the average.
©2018 AIC