Related Experiment Video
Updated: Jun 6, 2025
12:06
Analyzing Mitochondrial Morphology Through Simulation Supervised Learning
Published on: March 3, 2023
3.9K
Deep learning-enhanced automated mitochondrial segmentation in FIB-SEM images using an entropy-weighted ensemble
Yubraj Gupta1,2, Rainer Heintzmann2,3, Carlos Costa1
1Departamento de Electrónica, Telecomunicações e Informática (DETI), University of Aveiro, Aveiro, Portugal.
Plos One
|November 26, 2024
Summary
Automated segmentation of mitochondria using an ensemble of pipelines aids in early disease detection. This method efficiently processes large imaging datasets, improving accuracy for mitochondrial disease diagnosis.
Area of Science:
- Cell Biology
- Medical Imaging
- Computational Pathology
Background:
- Mitochondria are vital organelles producing cellular energy (ATP) via nutrient breakdown.
- Mitochondrial DNA alterations are linked to primary mitochondrial diseases, including neurodegenerative disorders.
- Early detection of mitochondrial abnormalities is critical for disease management.
Purpose of the Study:
- To develop an automated tool for efficient mitochondrial segmentation in large-scale microscopy datasets.
- To address the time-consuming nature of manual segmentation for pathologists.
- To improve the early detection and mitigation of mitochondrial diseases.
Main Methods:
- Proposed an ensemble of two automatic segmentation pipelines for mitochondria.
- Utilized an entropy-weighted fusion technique to combine pipeline outputs.
- Evaluated performance using Jaccard index, Dice coefficient, error rate, and pixel accuracy on four public datasets.
Main Results:
- The ensemble method achieved high performance across four datasets, with mean Jaccard index of 0.9644 and Dice coefficient of 0.9749 on the UroCell dataset.
- Demonstrated superior segmentation efficiency compared to 2D and 3D Convolutional Neural Network (CNN) algorithms.
- Achieved a mean error rate of 0.0062 and pixel accuracy of 0.9938.
Conclusions:
- The proposed automated ensemble segmentation approach offers efficient and accurate mitochondria identification.
- This method has the potential to significantly aid pathologists in the early diagnosis of mitochondrial diseases.
- The technique minimizes user intervention, making large-scale image analysis more feasible.
