Quick start

This example reuses the supplied three-channel masks, builds meshes, curates the cells, detects GFP-DnaN objects, and saves morphology and signal measurements. The script first copies the dataset so that the original inputs and outputs remain available for comparison.

Run this in the BactoScoop environment, or in its Jupyter kernel:

"""Three-channel analysis from downloaded examples, with supplied masks."""
from datetime import datetime
import json
import os
from pathlib import Path
import shutil
import bactoscoop

examples = os.environ.get('BACTOSCOOP_EXAMPLES_DIR')
candidates = [Path(examples)] if examples else [candidate for base in (Path.cwd(), *Path.cwd().parents) for candidate in (base/'examples',base)]
source = next((base/'3_channel_example' for base in candidates if (base/'3_channel_example').is_dir()),None)
if source is None:
    raise FileNotFoundError('Download the release examples and run from that folder, or set BACTOSCOOP_EXAMPLES_DIR.')
destination = Path.home()/'bactoscoop_runs'/('3_channel_example_'+datetime.now().strftime('%Y%m%d_%H%M%S_%f'))
shutil.copytree(source,destination)
ic = bactoscoop.ImageCollection(str(destination),px=0.065)
ic.create_image_objects(phase_channel='C1')
ic.batch_process_mesh(phase_channel='C1',join_thresh=4,split_thresh=0.5,CD_width=False,smoothing=0.1,save_data=True)
ic.curate_dataset(str(destination/'DnaN_Timecourse.pkl'),control=False)
ic.batch_detect_objects(channels=['C3'],reset_channels=True,align=False,smoothing=0.1,log_sigma=3,kernel_width=3,min_overlap_ratio=0.001,max_external_ratio=0.3)
ic.load_channel_images(['C2'])
ic.add_channels(ic.image_objects,['C2'],load_data=False)
ic.batch_calculate_features([([None],'morphological'),(['C2','C3'],'profiling'),(['C3'],'objects')],all_data=False,reset=True,max_mesh_size=1000)
table = ic.merge_dataframes(discard_morphological_nan=True)
ic.dataframe_to_parquet()
ic.dataframe_to_pkl()
(destination/'processing_summary.json').write_text(json.dumps(ic.processing_summary(),indent=2),encoding='utf-8')
print(f'{len(table)} retained rows; outputs in {destination}')

Download the complete script. The default calibration matches the supplied walkthrough; replace it with your microscope’s µm/pixel calibration for your own images.

What to inspect

Inspect the mask outlines before trusting the reconstructed meshes. After curation, compare the kept and rejected contours. Then inspect individual object detections and check whether the merged table contains the expected channels. See Plotting and visual quality control for the plotting functions.

This quick start intentionally uses the supplied SVM with its matching dataset. The same model need not classify cells correctly in a different experiment. Review SVM curation before applying it to other images.

Expected files

The working copy contains new _meshdata.pkl, _curated_meshdata.pkl, _features.pkl, and _features.parquet files. The script also saves processing_summary.json so item-level and cell-level failures can be reviewed. Existing masks remain in the working copy’s masks/ folder.