Skip to content
This repository has been archived by the owner on Oct 30, 2024. It is now read-only.

Commit

Permalink
Analyse data from experiments run on multiple days (#57)
Browse files Browse the repository at this point in the history 
* Add code to pool directions

* Add notes

* Cleanup

* Fix broken tests

* Recalculate outputs

* Update manifest and pyproject

* Fix package name

* Update package links

* Update dependencies

* Add more descriptions

* Edit pyproject file

* Expand readme with more details on the analysis

* Add docstring

* Test if we can see the schema in the readme file

* Add handling of multiple days

* Improve performance to deal with very large files, from O(n ^2) to O(1)

* Delete commented code

* Handle fit fails

* Save the data in a given folder

* Skip ROIs that do not fit

* Add missing fit handling in sftf analysis

* Regenerate mock data according to new "day" values

* Update docstring

* Remove big chunk of duplicated code

* Test image in readme file

* Remove image

* Apply suggestions from code review

Co-authored-by: Niko Sirmpilatze <[email protected]>

* Change prompt message

* Adjust input params of method get_median_subtracted_response_and_params

* Renaming and other small suggested fixes

* Renaming 2

* Small fixes

* Move out the method to make gaussian matrices

* Clean up

* Add tests for multiple days and add more structure and comments regarding creation of mock data

* Update readme

* Small updates in the docs

* Simplify fitting methods according to discussion

* Improve description on dataset creation optimization

* Remove unused config

* Rearrange methods in generate mock data

* Renaming ad adding new fixture

* Add missing n_roi fixture calls

* Renaming and folder reorganization

* Remove try-catch construct

* Linting

* Add missing adjustments

* Add readme file with a description of the regression tests

* Add more descriptions / docs

* Revert readme change

* Fix usage of conftest

* Linting

---------

Co-authored-by: Niko Sirmpilatze <[email protected]>
  • Loading branch information
@lauraporta @niksirbi
lauraporta and niksirbi authored Jun 15, 2023
1 parent ccaf866 commit 5728e8f
Show file tree
Hide file tree
Showing 13 changed files with 783 additions and 383 deletions.
139 changes: 78 additions & 61 deletions rsp_vision/analysis/spatial_freq_temporal_freq.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@
import numpy as np
import pandas as pd
import scipy.stats as ss
from scipy.optimize import OptimizeResult

from rsp_vision.analysis.gaussians_calculations import (
fit_2D_gaussian_to_data,
Expand Down Expand Up @@ -262,7 +263,8 @@ def nonparam_anova_over_rois(self) -> dict:
samples = np.zeros(
(
len(self.data.directions)
* self.data.n_triggers_per_stimulus,
* self.data.n_triggers_per_stimulus
* self.data.total_n_days,
len(self.data.sf_tf_combinations),
)
)
Expand Down Expand Up @@ -525,72 +527,80 @@ def get_median_subtracted_response_and_params(
return median_subtracted_response_2d_matrix, sf_0, tf_0, peak_response

def get_gaussian_fits_for_roi(self, roi_id: int) -> dict:
"""Calculate the best fit parameters for each ROI and direction.
This method takes as input the ROI index, and loops over the
directions to calculate the best fit parameters for each spatial
and temporal frequency. First, it calls the
get_median_subtracted_response_and_params method to extract the
median subtracted response matrix for the given ROI and direction.
Then, the method performs a 2D Gaussian fit to the 2D response
matrix using the fit_2D_gaussian_to_data method. The resulting best
fit parameters are stored in a dictionary, where the keys are the
directions, and the values are tuples containing the
preferred spatial and temporal frequencies and the peak response
amplitude, the best fit parameter values obtained from the Gaussian
fit, and the median-subtracted response matrix.
Args:
roi_id (int): The index of the ROI for which to calculate the best
fit parameters.
"""Calculates the best fit parameters for each direction and for
the pooled data for a given ROI. It calls the `manage_fitting`
method in oredr to find the best fit parameters.
Returns:
dict: A dictionary with the best fit parameters for each direction.
The keys are the directions, and the values are tuples
containing the preferred spatial and temporal frequencies
and the peak response amplitude, the best fit parameter values
obtained from the Gaussian fit, and the median-subtracted
response matrix.
Parameters
----------
roi_id (int)
The index of the ROI for which to calculate the best
fit parameters.
Returns
-------
dict
A dictionary with the best fit parameters for each direction
and for the pooled data.The keys are the directions, and the
values are tuples containing the preferred spatial and temporal
frequencies and the peak response amplitude, the best fit parameter
values obtained from the Gaussian fit, and the median-subtracted
response matrix.
"""

roi_data = {}
for dir in self.data.directions:
(
response_matrix,
sf_0,
tf_0,
peak_response,
) = self.get_median_subtracted_response_and_params(
responses=self.data.responses,
roi_data[dir] = self.manage_fitting(
roi_id=roi_id,
sfs=self.data.spatial_frequencies,
tfs=self.data.temporal_frequencies,
dir=dir,
direction=dir,
pool_directions=False,
)

initial_parameters = [
peak_response,
sf_0,
tf_0,
np.std(self.data.spatial_frequencies, ddof=1),
np.std(self.data.temporal_frequencies, ddof=1),
self.data.config["fitting"]["power_law_exp"],
]
# now the same by pooling directions
roi_data["pooled"] = self.manage_fitting(
roi_id=roi_id,
pool_directions=True,
)

best_result = fit_2D_gaussian_to_data(
self.data.spatial_frequencies,
self.data.temporal_frequencies,
response_matrix,
initial_parameters,
self.data.config,
)
return roi_data

roi_data[dir] = (
(sf_0, tf_0, peak_response),
best_result.x,
response_matrix,
)
def manage_fitting(
self,
roi_id: int,
pool_directions: bool = False,
direction: float = sys.float_info.min,
) -> Tuple[Tuple[float, float, float], np.ndarray, np.ndarray]:
"""
This method is called by the get_gaussian_fits_for_roi method.
It calls the get_median_subtracted_response_and_params method to
extract the median subtracted response matrix for the given ROI and
direction. Then, it calls the fit_2D_gaussian_to_data method to
perform a 2D Gaussian fit to the 2D response matrix. The resulting
best fit parameters are stored in a tuple, where the first element
is the peak response amplitude, the second element is the best fit
parameter values obtained from the Gaussian fit, and the third
element is the median-subtracted response matrix.
# now the same by pooling directions
Parameters
----------
roi_id : int
The ID of the ROI to extract the response matrix from.
pool_directions : bool, optional
Whether to extract the response matrix for a single direction,
by default False
direction : float, optional
The direction to extract the response matrix from. To be used
only if `pool_directions` is set to False. By default
sys.float_info.min
Returns
-------
Tuple[Tuple[float, float, float], np.ndarray, np.ndarray]
A tuple containing the peak response amplitude and its
corresponding SF and TF, the best fit parameter values
obtained from the Gaussian fit, and the median-subtracted
response matrix.
"""
(
response_matrix,
sf_0,
Expand All @@ -601,7 +611,8 @@ def get_gaussian_fits_for_roi(self, roi_id: int) -> dict:
roi_id=roi_id,
sfs=self.data.spatial_frequencies,
tfs=self.data.temporal_frequencies,
pool_directions=True,
dir=direction,
pool_directions=pool_directions,
)

initial_parameters = [
Expand All @@ -621,14 +632,20 @@ def get_gaussian_fits_for_roi(self, roi_id: int) -> dict:
self.data.config,
)

roi_data["pooled"] = (
if best_result is None:
logging.warning(
f"ROI {roi_id} and direction {dir} failed to fit."
+ "Skipping..."
)
best_result = OptimizeResult()
best_result.x = np.nan * np.ones(6)

return (
(sf_0, tf_0, peak_response),
best_result.x,
response_matrix,
)

return roi_data

def get_all_fits(self) -> None:
"""Calculate the Gaussian fits for all ROIs using multiprocessing.
Expand Down
13 changes: 10 additions & 3 deletions rsp_vision/console_application/app.py
View file
Original file line number Diff line number Diff line change
@@ -1,6 +1,7 @@
import logging
import pickle
import sys
from pathlib import Path

import rich
from fancylog import fancylog
Expand Down Expand Up @@ -47,12 +48,15 @@ def analysis_pipeline() -> None:
# pipeline draft
start_logging()

# TODO: add TUI or GUI fuctionality to get input from user
folder_name = Prompt.ask(
" \
Please provide only the dataset name.\n \
Format: Mouse_Id_Hemisphere_BrainRegion_Monitor_position.\n \
Example: AK_1111739_hL_RSPd_monitor_front\n \
Example (1 day): AK_1111739_hL_RSPd_monitor_front\n \
Example (2 days): BY_IAA_1117276_hR_RSPg_monitor_front\n \
Example (1 day, BIG): AS_1112809_hL_V1_monitor_front-right_low\n \
Example (2 days, big file): CX_1112654_hL_RSPd_monitor_front\n \
Example (2 days, big file): CX_1112837_hL_RSPd_monitor_front\n \
📄"
)

Expand All @@ -71,7 +75,10 @@ def analysis_pipeline() -> None:
logging.info("Analysis finished")
logging.info(f"Updated photon_data object: {photon_data}")

with open(f"{folder_name}_data.pickle", "wb") as f:
saving_path = (
Path(config["paths"]["output"]) / f"{folder_name}_data.pickle"
)
with open(saving_path, "wb") as f:
pickle.dump(photon_data, f)
logging.info("Analysis saved")

Expand Down
Loading

0 comments on commit 5728e8f

Please sign in to comment.