Index ← Previous Topic Next Topic →

Dendritic Spine Classification

All detected spines are classified according to the currently selected classification scheme. NeuronStudio can classify spines according to a default classification scheme, or you may create and train your own classifier by defining output classes and providing examples of your spine types from actual data.

Default Classification Scheme

By default NeuronStudio clasifies spines into either Stubby, Thin, or Mushroom. This classification scheme makes use of each spine's head to neck diameter ratio, length to head diameter ratio, and head diameter to determine its proper type. To adjust the critical values that control this classification scheme, open the Spine Classifier dialog box from the Run menu and select the Default tab (see image below). Then adjust the following values:

Neck Ratio: Spines with head to neck diameter ratio greater than this value are considered Thin or Mushroom.

Thin Ratio: Spines that do not meet the Neck Ratio value and have a length to spine to head diameter above this value are classified as thin, otherwise as stubby.

Mushroom Size: Spines that meet the Neck Ratio value and have a head diameter equal or greater than this value are labeled as mushroom, otherwise as stubby.

The Reset button can be used to revert to the program's original values.

Default Classification Scheme

Trainable Classification Scheme

As an alternative to the default classifier provided by NeuronStudio, you may define your own trainnable classifier. This classification scheme makes use of an Artificial Neural Network (ANN) to learn how to classify spines using examples provided by you. To activate the trainable classifier use the Enable check box in the Trainable tab of the Spine Classifier dialog box (see image below). You can choose an existing classifier from the drop-down list or you can create a new one by clicking the New... button. Clicking the Delete button will delete the selected classifier as long as it is not currently active.

Trainable Classification Scheme

Clicking the New... button brings up the Create Spine Classifier dialog box (see image below). Here you can input a name for the classifier and its various output classes, separated by commas

New Trainable Classifier

To add examples into the current trainable classifier you first select the spines (see Selection and Deselection) and then choose Edit >> Add Selected Spines To Classifier from the main menu. You must set the spine type before you add them into the classifier (see Editing the Spines). Note that you may add them and edit their type afterwards from the Spine Classifier Dialog Box by selecting Edit... in the Trainable tab. This will bring up the Edit Trainning Examples dialog (see image below). In this dialog you can step through each example, and either delete it or reset its type.

Edit Trainable Classifier

Send questions or comments to: alfredo.rodriguez@mssm.edu