API Reference

SeqFISH_ADCG.fit_img_tiles — Function

fit_img_tiles(img,
              main_tile_width :: Int64,
              tile_overlap :: Int64,
              sigma_lb :: Float64,
              sigma_ub :: Float64,
              tau :: Float64,
              final_loss_improvement :: Float64,
              min_weight :: Float64,
              max_iters :: Int64,
              max_cd_iters :: Int64,
              noise_mean :: Float64,
              fit_alg :: AbstractString = "ADCG"
            )

Arguments:

img : a 2048x2048 image to fit
main_tile_width : the width of the main tile
tile_overlap : width of the overlaps of tiles with their neighbors
sigma_lb : the lowest allowed σ of a PSF
sigma_ub : the highest allowed σ of a PSF
tau : not used in current version
final_loss_improvement : ADCG terminates when the improvement in the loss function in subsequent iterations is less than this
min_weight : ADCG terminates when the next best PSF to add to the model has weight less than this
max_iters : The maximum number of points to add in a model of a tile
max_cd_iters : the maximum number of iterations of gradient descent to run after adding a PSF to the model to adjust the parameters of all PSFs in the model
noise_mean : the noise mean is subtracted from the image before fitting
fit_alg : 'ADCG' or 'DAO', uses the respective algorithm. Only ADCG is thorougly tested. Must be set as a keyword argument.

Returns:

DataFrame of points included in model at the end of the ADCG run
records object of points obtained at intermediate steps in the fitting process

Fits gaussian point spread functions on a square image of arbitrary sized by breaking it up into overlapping tiles of user specified width and overlap, then running ADCG on each tile and aggregating the results. It is necessary to call remove duplicates on the resultant image to remove the duplicates in the regions of overlapping tiles.

source

SeqFISH_ADCG.fit_2048x2048_img_tiles — Function

fit_2048x2048_img_tiles(img,
                        sigma_lb :: Float64,
                        sigma_ub :: Float64,
                        tau :: Float64,
                        final_loss_improvement :: Float64,
                        min_weight :: Float64,
                        max_iters :: Int64,
                        max_cd_iters :: Int64,
                        noise_mean :: Float64,
                        fit_alg :: AbstractString = "ADCG"
                        )

Arguments:

img : a 2048x2048 image to fit
sigma_lb : the lowest allowed σ of a PSF
sigma_ub : the highest allowed σ of a PSF
tau : not used in current version
final_loss_improvement : ADCG terminates when the improvement in the loss function in subsequent iterations is less than this
min_weight : ADCG terminates when the next best PSF to add to the model has weight less than this
max_iters : The maximum number of points to add in a model of a tile
max_cd_iters : the maximum number of iterations of gradient descent to run after adding a PSF to the model to adjust the parameters of all PSFs in the model
noise_mean : the noise mean is subtracted from the image before fitting
fit_alg : 'ADCG' or 'DAO', uses the respective algorithm. Only ADCG is thorougly tested. Must be set as a keyword argument.

Returns:

DataFrame of points included in model at the end of the ADCG run
records object of points obtained at intermediate steps in the fitting process

Fits gaussian point spread functions in a 2048x2048 pixel image with ADCG by splitting it into overlapping 70x70 pixel tiles.

source

SeqFISH_ADCG.remove_duplicates — Function

remove_duplicates(points :: DataFrame,
                img,
                sigma_lb :: Float64,
                sigma_ub :: Float64,
                tau :: Float64,
                noise_mean :: Float64,
                min_allowed_separation :: Float64,
                dims :: Int64 = 2
                )

Arguments:

img : a 2048x2048 image to fit
sigma_lb : the lowest allowed σ of a PSF
sigma_ub : the highest allowed σ of a PSF
tau : not used in current version
noise_mean : the noise mean is subtracted from the image before fitting
min_allowed_separation : Dots within this distance of each other are considered duplicates
dims : The dimension of the data

Returns:

DataFrame of points without duplicates.

Thins duplicates within min_allowed_separtion of each other from an image. Keeping only best dots such that all dots returned are at least min_allowed_separation from their nearest neighbor. This is necessary when fit by a tiled ADCG where the tiles overlap.

source

SeqFISH_ADCG.fit_tile — Function

fit_tile(inputs)

Arguments:

inputs : A tuple of inputs with entries:
- tile : An image tile to perform ADCG on.
- sigma_lb : the lowest allowed value of the width parameter of dots.
- sigma_ub : the highest allowed value of the width parameter of dots.
- noise_mean : estimated mean of noise. Pixel values below this are set to zero.
- tau : not used in this version
- final_loss_improvement : Terminate ADCG when the objective improves by less than this value in one iteration
- min_weight : The minimum allowed weight of a PSF in the image model
- max_iters : The maximum number of ADCG iterations, or number PSFs to add to the model.
- max_cd_iterations : the maximum number of times to perform gradient descent for the parameter values of all dots.
- fit_alg : 'ADCG' or 'DAO', uses the respective algorithm. Only ADCG is thorougly tested.

Returns:

records object of points obtained at intermediate steps in the fitting process

Runs ADCG on a single 2D tile.

source

SeqFISH_ADCG.fit_stack — Function

fit_stack(inputs)

Arguments:

inputs : A tuple of inputs with entries:
- stack : An image stack to perform ADCG on.
- sigma_xy_lb : the lowest allowed value of the width parameter of dots in the xy plane.
- sigma_xy_ub : the highest allowed value of the width parameter of dots in the xy plane.
- sigma_z_lb : the lowest allowed value of the width parameter of dots in the z axis.
- sigma_z_ub : the hightest allowed value of the width parameter of dots in the z axis.
- final_loss_improvement : Terminate ADCG when the objective improves by less than this value in one iteration
- min_weight : The minimum allowed weight of a PSF in the image model
- max_iters : The maximum number of ADCG iterations, or number PSFs to add to the model.
- max_cd_iterations : the maximum number of times to perform gradient descent for the parameter values of all dots.
- fit_alg : 'ADCG' or 'DAO', uses the respective algorithm. Only ADCG is thorougly tested. Must be set as a keyword argument.

Returns:

records object of points obtained at intermediate steps in the fitting process

Fits a model of a 3D image stack as a linear combination of point spread functions using ADCG.

source

SeqFISH_ADCG.fit_stack_tiles — Function

fit_stack_tiles(
                   img_stack,
                   main_tile_width :: Int64,
                   tile_overlap :: Int64,
                   tile_depth :: Int64,
                   tile_depth_overhang :: Int64,
                   sigma_xy_lb :: Float64,
                   sigma_xy_ub :: Float64,
                   sigma_z_lb :: Float64,
                   sigma_z_ub :: Float64,
                   final_loss_improvement :: Float64,
                   min_weight :: Float64,
                   max_iters :: Int64,
                   max_cd_iters :: Int64,
                   fit_alg = "ADCG"
            )

Arguments:

img : a 2048x2048 image to fit
main_tile_width : the width of the main tile
tile_overlap : width of the overlaps of tiles with their neighbors
tile_depth : the z depth of main tiles`
tile_depth_overhang : the overlap/overhang of tiles in the z axis
sigma_xy_lb : the lowest allowed lateral σ of a PSF
sigma_xy_ub : the highest allowed lateral σ of a PSF
sigma_z_lb : the lowest allowed axial σ of a PSF
sigma_z_ub : the highest allowed axial σ of a PSF
final_loss_improvement : ADCG terminates when the improvement in the loss function in subsequent iterations is less than this
min_weight : ADCG terminates when the next best PSF to add to the model has weight less than this
max_iters : the maximum number of iterations in which a new PSF may be added to the model of a tile (i.e. the maximum PSFs in a tile)
max_cd_iters : the maximum number of iterations of gradient descent to run after adding a PSF to the model to adjust the parameters of all PSFs in the model
fit_alg : 'ADCG' or 'DAO', uses the respective algorithm. Only ADCG is thorougly tested. Must be set as a keyword argument.

Returns:

DataFrame of points included in model at the end of the ADCG run
records object of points obtained at intermediate steps in the fitting process

Fits gaussian point spread functions using ADCG on a square image of arbitrary sized by breaking it up into overlapping tiles of user specified width and overlap, then running ADCG on each tile and aggregating the results. It is necessary to call remove duplicates on the resultant image to remove the duplicates in the regions of overlapping tiles.

source

SeqFISH_ADCG.get_mw_dots — Function

get_mw_dots(records :: DataFrame, w :: Real)

Look up the ADCG result at a minimum weight larger than the final minimum weight given for an ADCG run.

Returns a dataframe of results.

source