deepBlink

Threshold independent detection and localization of diffraction-limited spots.

Automatically detects spots given a 2D, 3D, or 2D+T image. Note that prediction is only possible in 2D and no NMS is performed in case 3D images are provided.

Requires that you have python 3 installed on your system (https://docs.knime.com/2020-12/python_installation_guide/index.html). Models can be downloaded through the command line interface using `deepblink download` or custom trained using `deepblink train`.

An example using this node can be found here (https://kni.me/w/OpqS9rYOZ5tL5iqZ).

Options

Perform Refinement: Check the box if a Gaussian-like profile should be fit over deepBlink's predicted spots. This generally does not improve performance but is more mathematically explainable.
Image Column: Column containing the image on which deepBlink should predict. Must be of type "IMG".
Refinement Radius (px): Pixel radius around deepBlink's detected spots to perform a Gaussian-like profile fitting for 10 iterations.
deepBlink Model: Path to the deepBlink model for spot detection. Can be a custom-trained model using `deepblink train` or a pre-trained model downloadable using `deepblink download`.

Input Ports

: Row with image on which should be predicted.

Output Ports

: Pass through of all input images to facilitate downstream processing.
: Table containing all detected spots with the following columns: - ("frame") slice/frame if a 3D or 2D+T image is provided - "spotID" as unique number for the each spot - "x, y" pixel coordinate location - "total_mass" is the total (sum) integrated brightness - "mean_mass" is the average (mean) brightness per pixel in the spot - "size" is the radius of gyration of its Gaussian-like profile - "ecc" is its eccentricity (where 0 is circular)

Nodes

Boolean Configuration1 ×
Column Selection Configuration1 ×
Component Input1 ×
Component Output1 ×
Conda Environment Propagation1 ×
Show all 12 nodes

deepBlink

Options

Input Ports

Output Ports

Nodes

Extensions

Links

Download