lenca56 / sbl Goto Github PK

Step 0: 
First, manually edit column names of the .csv files that come out of DLC such that the columns are now of the form "Frames","Ear1.X","Ear1.Y","Ear1.L","Ear2.X",etc.
Then run through "ExtractingOutliers&Expressions_Step0.ipynb" which mainly does some initial preprocessing and preanalysis of data
- this sets any particular frame with a bodypart of likelihood under 0.95 to missing values for all body parts
- checks how many missing value frames happen across stimuli intervals
- further extracts "outlier" frames by checking all frames for which a bodypart "jumps" a high distance between consecutive frames, then one can manually check in fiji if the jump is due to a true jump or a wrong label
- then dropping all likelihood columns "something.L"
- then creates intervals of interest for further analysis by importing stimulus start times and then subtracting a few seconds before and a few seconds after 
- defines generalized plot function for the facial expressions
- defines baseline expression by averaging facial expressions across a non-stimulus interval
- looks at what facial expressions looks like at the maximal Euclidian distance point from the baseline for each stimulus interval
Documents that should come out of this script for future scripts to work well, one of each of the 4 types for a single animal whose identity is given by batch number and animal number:
- "Batch" + str(batch) + "/OutputAnimal" + str(batch) + str(animall) + ".csv" which is dataframe with bodypart coordinates after Step0 processing
- "Batch" + str(batch) + "/list_start" + str(batch) + str(animall) + ".csv" which is the stimulus start time (in frame number)
- "Batch" + str(batch) + "/list_end" + str(batch) + str(animall) + ".csv" which is the end time of the stimulus interval of interest (in frame number)
- "Batch" + str(batch) + "/BaselineExpression" + str(batch) + str(animall) + ".csv" which is the facial expression X and Y coordinates for the avaraged baseline expression found in Step 0

Step 1:
Run through "FacialExpressionsMahalanobis_Step1.ipynb":
- mahalanobis_maximal(batch,animal) = function plotting facial expression of maximal Mahalanobis distance from baseline
- mahalanobis_all_peaks(batch,animal) = function looking at facial expression at any of the peaks in the Mahalanobis distance from baseline vs time plot
- prototype_mahalanobis(batch,animal) = function calculating prototypical facial expressions by averaging the maximal facial expression across all trials for a particular stimulus
- distance_to_baseline_averaged(batch,animal, stim) = function plotting Mahalanobis distance vs time averaged across all trials and showing standard deviation
- distance_to_baseline_averaged_all_animals(batch) = does the same thing as previous but now averaged across all animals, standard deviation across trials after averaged across all animals
- distance_to_baseline_averaged_all_animals_std(batch) = does same as previous but standard deviation is across animals, not trials
- mahalanobis_maximal_shock_two_intervals(batch, animall) = function looking at maximal facial expressions in two distinct shock intervals to try to distinguish an "early" vs a "late" shock behavioral response
- facial_expressions_trajectories_time(batch, animall) = function looking at progress of facial expresssion over time
- similarity_index(batch, animall) = function creating two "similarity" measures to compare any facial expression to a prototypical expression. one is pearson r correlation between facial expression and prototype, the other is Mahalanobis distance between the two. both are normalized by subtracting the value of the measure at the baseline facial expression


Step 2:
Run through "FacialExpressions_PolarCoordinates_Step2.ipynb":
- df_transform_polar(batch,animall) = function transforming all bodypart coordinates to polar coordinates (angle and distance), and normalize by subtracting baseline; 
it saves the converted dataframes as "Batch" + str(batch) + "/PolarCoordinates_Animal" + str(batch) + str(animall) + ".csv"
- ear_rotation_trials(batch, animall) = function plotting angle of earpoint Ear3 vs time for each trial/stimulus (that value represents how much the 3rd point of the ear, which is the top of the ear, rotates with respect to the initial baseline position)
- ear_rotation_whole_video(batch, animall) = functoin plotting the same as above but now across the whole video in order to check if there is a ear rotation tendency happening throughout the video due to increasing stress

Step 3:
Run through "LaplacianEigenMaps_Step3.ipynb":