dvid.tip.detect_tips

dvid.tip.detect_tips(x, use_clf=False, psd_dist=False, done_dist=False, checked_dist=50, tip_dist=False, pos_filter=None, save_to=None, verbose=True, snap=True, server=None, node=None)[source]

Detect potential open ends on a given neuron.

In brief, the workflow is as follows:

  1. Extract tips from the neuron’s skeleton

  2. Snap tip positions back to mesh (see snap)

  3. Apply filters (see _dist parameters)

  4. Remove tips close to a previously checked assignment (see checked_dist)

  5. (Optionally) Try to prioritize tips (use_clf)

  6. Save to a json file (see save_to) that can be opened in neuTu

Parameters

  • x (single body ID) –

  • use_clf (bool, optional) – If True, will use a pre-trained classifier to try to predict whether a tip needs human interaction or not. The returned list of tips will contain and be ordered by “confidence” values from -1 (unimportant) to +1 (important). THIS IS NOT A MAGIC BULLET! If you need to be certain you have completed a neuron, you will actually have to look at all tips regardless of confidence!

  • psd_dist (int | None, optional) – Minimum distance (in raw units) to a postsynaptic density (PSD) for a tip to be considered “done”.

  • done_dist (int | None,) – Minimum distance (in raw units) to a DONE tag for a tip to be considered “done”.

  • checked_dist (int | None, optional) – Minimum distance (in raw units) to a bookmark that has previously been “Set Checked” in the “Assigned bookmarks” table in Neutu.

  • tip_dist (int | None, optional) – If a given pair of tips is closer than this distance they will be considered duplicates and one of them will be dropped.

  • pos_filter (function, optional) – Function to fitler tips by position. Must accept numpy array (N, 3) and return array of [True, False, …].

  • save_to (filepath, optional) – If provided will save open ends to JSON file that can be imported as assigmnents.

  • snap (bool, optional) – If True, will make sure that tips positions are within the mesh.

  • server (str, optional) – If not provided, will try reading from global.

  • node (str, optional) – If not provided, will try reading from global.

Returns

List of potential open ends.

Return type

pandas.DataFrame

Examples

The usual setting up:

>>> import dvidtools as dt
>>> dt.set_param('https://your.server.com:8000', 'node', 'user')

By default, detect_tips will simply return all end nodes of the skeleton that have not previously been “Set checked” in NeuTu:

>>> # Generate list of tips and save to json file
>>> tips = dt.detect_tips(883338122, save_to='~/Documents/883338122.json')

A more advanced example is using a pre-trained classifier to include “confidence” values. These confidence range from -1 to +1 and give some indication whether a tip needs to be extended or not. This requires sciki-learn to be installed. In a terminal run:

pip install scikit-learn

Once scikit-learn is installed, you can run the tip detector with classifier confidences:

>>> tips = dt.detect_tips(883338122, use_clf=True,
...                      save_to='~/Documents/883338122.json')