Detect - lucadellasantina/ObjectFinder GitHub Wiki

Detect objects is the main function of ObjectFinder.

Main toolbar (top section)

Create New Project: Select the Location, the Images and Masks to be analysed for a new project.

Open Project: Select an experiment folder to perform the analysis. Valid folders must contain a subfolder called "I" in which all the image stacks will be stored. Allowed images must be 8-bit .tif files.

Choose settings: Respond a few questions about the images to be analyzed for ObjectFinder to determine the best settings for it.

Detect Objects: This is the main function of ObjectFinder, candidate objects will be searched within the present image volume according to the settings below using the masks and skeletons selected.

Validate Objects: Once found candidate objects, the user is asked to validate them using one of these approaches:

  1. Sliced Volume Inspector (defualt): inspect visually slice by slice and validate objects. The validation may be done manually selecting each object or many objects at the same time by encircling them using the lasso tool. Additionally, two filters may be added to exclude objects by Score/Volume/Brightness/Roundness/Major Axis Length/Z position. Validation changes may be saved onto the current set of objects or onto a new set.
  2. Volume Inspector: inspect visually the 3D rendered volume and validate objects. Validation may be done selecting thresholds of object statistics such as Score/Volume/Brightness/Roundness/Major Axis Length/Z position. This visualization mode can be slow for large objects set and for computers with limited RAM and GPU specs.
  3. Bitplane Imaris: inspect visually objects by rendering them within Imaris, validate objects either manually by interacting with them or by defining filters in Imaris of the objects properties (e.g. Volume, Score, Avg Intensity etc...)
  4. Neural Net: automatically validate objects using a deep-learning classifier trained by the user

Combo boxes (middle section)

Below the main toolbar, the following five combo-boxes will allow you to define players of the object detection process:

  1. Objects: Shows objects previously found in the selected image for this project. Objects may be: Renamed/Copied/Erased/Visualized as a map of their position

  2. Images: Shows available image stacks. Select which image will be analysed among the ones saved within the "I" folder. You can check: Image Properties/Render the image/Display a maximum intensity projection/Choose from the gallery fro this project.

  3. Masks: Shows available image stacks. Select a binary mask to restrict search operations when detecting objects on this project. You can check: Image Properties/Render the masked volume/Display a maximum intensity projection of the masked image/Display a maximum intensity projection of the masked image.

  4. Skeletons: Select a skeleton to be used when analyzing objects on this project. Skeletons may be: Renamed/Imported from file/Erased/Plotted

  5. Validation method: Select among the available methods which one to use to discriminate between valid and invalid objects among detected candidates. Invalid objects may be purged from the current object set here.

Object Detection Settings (bottom section)

This section will allow you to select custom settings for the detection process:

  • Load/Save settings: Previous settings may be loaded onto any project from file. Current settings may be saved.
  • Objects diameter range: minimum and maximum allowed object diameter within the image volume. The calculator might help to determine the best settings for the objects you are expecting to detect.
  • Detection Algorithm: Iterative thresholding (default) is the most accurate algorithm, Local thresholding is less accurate but faster. Use the latter only if expecting to detect trillions of objects.
  • Voxel Connectivity: Defines in how many directions ObjectFinder should look for connected voxels in 3D, available options are 6 (ObjectFinder default, common face), 18 (common face/edge), 26 (ImageJ/Fiji default, common face/edge/corner).
  • Watershed: Split candidate objects with multiple intensity peak as individual objects using watershed segmentation.
  • Block search: Search the entire volume by splitting into smaller sub-volumes. This option allows calculation of local background level when enabled, as opposed to calculating a single background level for the entire image volume.
  • Local Background: Select between local backgrounds in block search or a single background for the entire image
  • Shape properties: Calculate shape properties of objects such as oblongness and major axis length.
  • Background level: Select how background level is detected among Most common intensity/Minimum Intensity/Standard Deviation
  • Minimum Object Intensity: number of times above background level upon which to detect candidate objects
  • Filter objects on a single Z plane: remove all candidate objects that are spanning only across a single Z plane
  • Filter objects on the mask's edge: remove all candidate objects that are touching the edge of the search mask
  • Filter objects with unstable centroids: remove objects if their center of mass is not stable along Z