Mercurial > repos > bgruening > cp_export_to_spreadsheet
diff track_objects.py @ 4:f8fa0f6718a3 draft
"planemo upload for repository https://github.com/bgruening/galaxytools/tree/master/tools commit 35da2dcd86747c9bff138e100dbe08c6106f3780"
author | bgruening |
---|---|
date | Sat, 06 Feb 2021 10:07:37 +0000 |
parents | |
children | d0178bdcd00e |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/track_objects.py Sat Feb 06 10:07:37 2021 +0000 @@ -0,0 +1,199 @@ +#!/usr/bin/env python + +import argparse +import json + +from cp_common_functions import get_json_value +from cp_common_functions import get_pipeline_lines +from cp_common_functions import get_total_number_of_modules +from cp_common_functions import INDENTATION +from cp_common_functions import update_module_count +from cp_common_functions import write_pipeline + +MODULE_NAME = "TrackObjects" +OUTPUT_FILENAME = "output.cppipe" + + +def build_header(module_name, module_number): + result = "|".join([f"{module_name}:[module_num:{module_number}", + "svn_version:\\'Unknown\\'", + "variable_revision_number:7", + "show_window:True", + "notes:\\x5B\\'Track the embryos across images using the Overlap method\\x3A tracked objects are identified by the amount of frame-to-frame overlap. Save an image of embryos labeled with a unique number across time.\\'\\x5D", + "batch_state:array(\\x5B\\x5D, dtype=uint8)", + "enabled:True", + "wants_pause:False]\n"]) + return result + + +def build_main_block(input_params): + result = INDENTATION.join([f"{INDENTATION}Choose a tracking method:{get_json_value(input_params,'con_tracking_method.tracking_method')}\n", + f"Select the objects to track:{get_json_value(input_params,'object_to_track')}\n" + ]) + + tracking_method = get_json_value(input_params, 'con_tracking_method.tracking_method') + + obj_measurement = "None" # default value + if tracking_method == "Measurements": + measurement_category = get_json_value(input_params, 'con_tracking_method.con_measurement_category.measurement_category') + measurement = get_json_value(input_params, 'con_tracking_method.con_measurement_category.measurement') + + if measurement_category == "Intensity" or measurement_category == "Location": + img_measure = get_json_value(input_params, 'con_tracking_method.con_measurement_category.img_measure') + obj_measurement = f"{measurement_category}_{measurement}_{img_measure}" + else: + obj_measurement = f"{measurement_category}_{measurement}" + + result += INDENTATION.join([f"{INDENTATION}Select object measurement to use for tracking:{obj_measurement}\n"]) + + if tracking_method == "LAP": # no max distance required, set default for pipeline + max_distance = 50 + else: + max_distance = get_json_value(input_params, 'con_tracking_method.max_distance') + + result += INDENTATION.join([f"{INDENTATION}Maximum pixel distance to consider matches:{max_distance}\n"]) + + display_option = get_json_value(input_params, 'con_tracking_method.display_option') + + output_img_name = "TrackedCells" # default value, required by cppipe regardless of its presence in UI + save = get_json_value(input_params, 'con_tracking_method.con_save_coded_img.save_coded_img') + if save == "Yes": + output_img_name = get_json_value(input_params, 'con_tracking_method.con_save_coded_img.name_output_img') + + result += INDENTATION.join( + [f"{INDENTATION}Select display option:{display_option}\n", + f"Save color-coded image?:{save}\n", + f"Name the output image:{output_img_name}\n" + ]) + + # LAP method default values + movement_model = "Both" + no_std = 3.0 + radius_limit_max = 10.0 + radius_limit_min = 2.0 + radius = "2.0,10.0" + run_second = "Yes" + gap_closing = 40 + split_alt = 40 + merge_alt = 40 + max_gap_displacement = 5 + max_split = 50 + max_merge = 50 + max_temporal = 5 + max_mitosis_dist = 40 + mitosis_alt = 80 + + # LAP method + if tracking_method == "LAP": + movement_model = get_json_value(input_params, 'con_tracking_method.movement_method') + no_std = get_json_value(input_params, 'con_tracking_method.no_std_radius') + radius_limit_max = get_json_value(input_params, 'con_tracking_method.max_radius') + radius_limit_min = get_json_value(input_params, 'con_tracking_method.min_radius') + radius = f"{radius_limit_min},{radius_limit_max}" + + run_second = get_json_value(input_params, 'con_tracking_method.con_second_lap.second_lap') + if run_second == "Yes": + gap_closing = get_json_value(input_params, 'con_tracking_method.con_second_lap.gap_closing') + split_alt = get_json_value(input_params, 'con_tracking_method.con_second_lap.split_alt') + merge_alt = get_json_value(input_params, 'con_tracking_method.con_second_lap.merge_alt') + max_gap_displacement = get_json_value(input_params, 'con_tracking_method.con_second_lap.max_gap_displacement') + max_split = get_json_value(input_params, 'con_tracking_method.con_second_lap.max_split') + max_merge = get_json_value(input_params, 'con_tracking_method.con_second_lap.max_merge') + max_temporal = get_json_value(input_params, 'con_tracking_method.con_second_lap.max_temporal') + max_mitosis_dist = get_json_value(input_params, 'con_tracking_method.con_second_lap.max_mitosis_distance') + mitosis_alt = get_json_value(input_params, 'con_tracking_method.con_second_lap.mitosis_alt') + + result += INDENTATION.join( + [f"{INDENTATION}Select the movement model:{movement_model}\n", + f"Number of standard deviations for search radius:{no_std}\n", + f"Search radius limit, in pixel units (Min,Max):{radius}\n", + f"Run the second phase of the LAP algorithm?:{run_second}\n", + f"Gap closing cost:{gap_closing}\n", + f"Split alternative cost:{split_alt}\n", + f"Merge alternative cost:{merge_alt}\n", + f"Maximum gap displacement, in pixel units:{max_gap_displacement}\n", + f"Maximum split score:{max_split}\n", + f"Maximum merge score:{max_merge}\n", + f"Maximum temporal gap, in frames:{max_temporal}\n" + ]) + + # common section + filter_by_lifetime = get_json_value(input_params, 'con_tracking_method.con_filter_by_lifetime.filter_by_lifetime') + use_min = "Yes" # default + min_life = 1 # default + use_max = "No" # default + max_life = 100 # default + + if filter_by_lifetime == "Yes": + use_min = get_json_value(input_params, 'con_tracking_method.con_filter_by_lifetime.con_use_min.use_min') + if use_min == "Yes": + min_life = get_json_value(input_params, 'con_tracking_method.con_filter_by_lifetime.con_use_min.min_lifetime') + + use_max = get_json_value(input_params, 'con_tracking_method.con_filter_by_lifetime.con_use_max.use_max') + if use_max == "Yes": + max_life = get_json_value(input_params, 'con_tracking_method.con_filter_by_lifetime.con_use_max.max_lifetime') + + result += INDENTATION.join( + [f"{INDENTATION}Filter objects by lifetime?:{filter_by_lifetime}\n", + f"Filter using a minimum lifetime?:{use_min}\n", + f"Minimum lifetime:{min_life}\n", + f"Filter using a maximum lifetime?:{use_max}\n", + f"Maximum lifetime:{max_life}\n" + ]) + + # print 2 leftover from LAP + result += INDENTATION.join( + [f"{INDENTATION}Mitosis alternative cost:{mitosis_alt}\n", + f"Maximum mitosis distance, in pixel units:{max_mitosis_dist}\n" + ]) + + # Follow Neighbors + # defaults + avg_cell_diameter = 35.0 + use_adv = "No" + cost_of_cell = 15.0 + weight_of_area_diff = 25.0 + + if tracking_method == "Follow Neighbors": + avg_cell_diameter = get_json_value(input_params, 'con_tracking_method.avg_diameter') + use_adv = get_json_value(input_params, 'con_tracking_method.con_adv_parameter.adv_parameter') + if use_adv == "Yes": + cost_of_cell = get_json_value(input_params, 'con_tracking_method.con_adv_parameter.cost') + weight_of_area_diff = get_json_value(input_params, 'con_tracking_method.con_adv_parameter.weight') + + result += INDENTATION.join( + [f"{INDENTATION}Average cell diameter in pixels:{avg_cell_diameter}\n", + f"Use advanced configuration parameters:{use_adv}\n", + f"Cost of cell to empty matching:{cost_of_cell}\n", + f"Weight of area difference in function matching cost:{weight_of_area_diff}\n" + ]) + + return result + + +if __name__ == "__main__": + parser = argparse.ArgumentParser() + parser.add_argument( + '-p', '--pipeline', + help='CellProfiler pipeline' + ) + parser.add_argument( + '-i', '--inputs', + help='JSON inputs from Galaxy' + ) + args = parser.parse_args() + + pipeline_lines = get_pipeline_lines(args.pipeline) + inputs_galaxy = json.load(open(args.inputs, "r")) + + current_module_num = get_total_number_of_modules(pipeline_lines) + current_module_num += 1 + pipeline_lines = update_module_count(pipeline_lines, current_module_num) + + header_block = build_header(MODULE_NAME, current_module_num) + main_block = build_main_block(inputs_galaxy) + + module_pipeline = f"\n{header_block}{main_block}\n" + pipeline_lines.append(module_pipeline) + + write_pipeline(OUTPUT_FILENAME, pipeline_lines)