Automatic detection of three cell types in a microscope image based on deep learning.

Area of Science:

• Computational Biology
• Medical Imaging
• Deep Learning

Background:

• Accurate cell detection and classification are crucial for medical image analysis.
• Existing target detection models face challenges with complex cellular structures and varying image scales.

Purpose of the Study:

• To develop an advanced deep learning model for precise human cell detection and classification.
• To enhance multi-scale feature fusion for improved performance in cell image analysis.

Main Methods:

• A novel feature extraction network using a residual neural network with Instance Normalization and Mish activation.
• A new feature fusion network integrating multi-scale feature graphs.
• Gaussian hybrid clustering for hyperparameter optimization.

Main Results:

• The proposed model achieved an average accuracy exceeding 0.96 on a human cell image dataset.
• Demonstrated an 11.9% improvement in accuracy compared to existing target detection methods.
• Showcased adaptability to datasets with uneven sample distributions.

Conclusions:

• The developed model offers a significant advancement in cell image analysis.
• Provides a robust solution for accurate cell detection and classification in medical imaging.
• Offers new research avenues for medical image analysis, particularly for imbalanced datasets.