Mito-Analysis

Mitochondria Segmentation, Skeletonization, and Feature Extraction Pipeline

This repository contains a complete analysis pipeline for automated segmentation, skeletonization, and quantitative feature extraction of mitochondria from wild-type (WT) and knockout (KO) samples. The project focuses on identifying morphological and textural heterogeneity using a deep-learning–based segmentation model followed by automated skeleton extraction and quantitative profiling.

---

1. Project Overview

Mitochondria exhibit large structural variability across different knockout types. This pipeline enables:

1. Deep learning–based mitochondrial segmentation using a trained U-Net model.
2. Automated skeletonization to extract cristae structure using Fiji.
3. Measurement of intensity, shape, morphology, and texture features.
4. Export of standardized CSV outputs for downstream analysis.

---

2. Pipeline Components

A. Deep Learning Segmentation (Python)

• Uses a U-Net model to segment mitochondria despite morphological variability.
• Differentiates hollow and textured mitochondrial phenotypes.
• Outputs binary masks and visual overlays.

B. Skeletonization (python + Fiji Macro)

• Skeletonizes mitochondria masks.
• Generates ROIs corresponding to cristae paths.
• Profiles intensity along skeletons using NeuroCyto plugin tools.
• Exports ROIs, tables, and skeleton images.

C. Feature Extraction

Combines information from:

• Raw images
• Segmentation masks
• Skeletons and intensity profiles

Outputs include:

• Morphological features
• Intensity features
• Texture features (e.g., GLCM)
• Combined CSV files

---

3. Installation and Environment Setup

A. Python Environment

Install required Python packages:

pip install numpy scipy pandas scikit-image requests tifffile connected-components-3d
pip install torch nnunetv2

Note: For nnU-Net installation details, GPU requirements, and troubleshooting, see: https://github.com/MIC-DKFZ/nnUNet

Note: For pytorch with cuda installation details, GPU requirements, and troubleshooting, see: https://pytorch.org/get-started/locally/

---

B. Fiji / ImageJ Setup

Install Fiji: https://imagej.net/software/fiji/

Plugin req : NeuroCyto Toolbox