FiberSight
What is FiberSight? Link to heading
FiberSight is software I designed to bridge the gap between open-source imaging, machine learning, and muscle image analysis. It was built as a plugin to FIJI/ImageJ, but with functionality driven by machine-learning from Cellpose via Python. Multiple deep-learning models were finetuned according to specific typical stainings of muscle fibers borders (PSR, H&E, and WGA) in order to produce accurate segmentations of whole-slide images.

Launcher from FIJI/ImageJ
What can FiberSight do? Link to heading
FiberSight combines 3 processes into a single pipeline:
- Muscle Fiber Segmentation (via custom Cellpose)

Segmentation of PSR-Stained Muscle Fibers
-
Muscle Fiber Nucleation-State (via ROI erosion)
Muscle Fibers co-stained with DAPI and Laminin. Red = More centrally nucleated, Green = More Peripherally Nucleated
-
Muscle Fiber Type (via Thresholding and Custom Nearest-Neighbor Determination)
Immunostained Fluorescence Muscle Fiber-Typing
Finally, FiberSight combines all of these operations to produce and save a single spreadsheet for each image, quantifying each fiber by its detected morphological and phenotypic properties

Results combined into a spreadsheet
Check out the FiberSight Project on GitHub for more details on usage and installation!
The typical operation of FiberSight is below: