Fast Multi-Focus Fusion Based on Deep Learning for Early-Stage Embryo Image Enhancement
Vidas Raudonis1, Agne Paulauskaite-Taraseviciene2, Kristina Sutiene3
1Department of Automation, Kaunas University of Technology, Studentu 48, 51367 Kaunas, Lithuania.
Sensors (Basel, Switzerland)
|February 2, 2021
Summary
This study introduces a deep learning U-Net based multi-focus image fusion method for improved early-stage embryo cell detection and counting. The approach significantly reduces data size while maintaining spectral information for enhanced microscopic imaging quality.
Area of Science:
- Embryology
- Medical Imaging
- Computational Biology
Background:
- Accurate cell detection and counting are crucial for assessing early-stage embryo quality.
- Automation is challenging due to variations in cell morphology, overlapping cells, and image quality.
- Efficient processing of large datasets is required for clinical applications.
Purpose of the Study:
- To develop an automated method for cell detection and counting in early-stage embryos.
- To enhance the quality of microscopic embryo images for better analysis.
- To reduce data processing time and storage requirements.
Main Methods:
- A multi-focus image fusion technique utilizing a deep learning U-Net architecture.
- Data reduction by up to 7 times without loss of essential spectral information.
- Comparative analysis with Inverse Laplacian Pyramid Transform and Enhanced Correlation Coefficient Maximization.
Main Results:
- Substantial improvement in image fusion time across various resolutions.
- High quality of the fused images maintained.
- Demonstrated efficiency compared to existing methods.
Conclusions:
- The proposed U-Net based image fusion method offers significant advantages in processing time.
- The technique ensures high-quality fused images essential for accurate embryo evaluation.
- This approach addresses key challenges in automated embryo analysis.
