Comprehensive data analysis of white blood cells with classification and segmentation by using deep learning...

Area of Science:

• Medical Image Analysis
• Computational Biology
• Artificial Intelligence in Medicine

Background:

• Deep learning is widely used for human peripheral blood cell analysis.
• Previous studies often analyzed datasets separately, limiting generalizability.
• Combining multiple datasets for blood cell classification and segmentation remains underexplored.

Purpose of the Study:

• To develop and evaluate a deep learning model for classifying and segmenting human peripheral blood cells.
• To investigate the efficacy of combining multiple datasets for improved model performance.
• To assess the model's accuracy on train-independent datasets for clinical applicability.

Main Methods:

• Utilized a mixture of four distinct datasets for white blood cell classification.
• Applied three neural network architectures (CNN, UNet, SegNet) for segmentation.
• Proposed a Convolutional Neural Network (CNN) for nucleus and cytoplasm detection.

Main Results:

• Achieved 98.03% balanced accuracy for classification and 97.27% test accuracy on a train-independent dataset.
• The proposed CNN achieved 98.9% accuracy on a train-dependent dataset and 92.82% on a train-independent dataset for segmentation.
• Demonstrated high accuracy in detecting white blood cells from train-independent datasets.

Conclusions:

• The proposed deep learning method effectively classifies and segments white blood cells using combined datasets.
• The model exhibits strong performance on unseen data, indicating robustness.
• The approach shows significant promise as a diagnostic tool for clinical applications.