White blood cell classification via a discriminative region detection assisted feature aggregation network

Area of Science:

• Medical Imaging Analysis
• Computational Pathology
• Artificial Intelligence in Diagnostics

Background:

• Accurate white blood cell (WBC) classification is crucial for pathological diagnosis.
• Existing methods face challenges like uneven staining, boundary blur, and nuclear variability.
• These issues limit classification accuracy in real-world scenarios.

Purpose of the Study:

• To propose a novel deep neural network, DRFA-Net, for improved WBC classification.
• To enhance classification performance by accurately locating WBC areas and aggregating discriminative features.
• To address practical limitations affecting current WBC classification models.

Main Methods:

• Developed DRFA-Net, a deep neural network incorporating discriminative region detection and feature aggregation.
• Implemented an adaptive feature enhancement module for refining multi-level deep features.
• Designed a network branch for WBC area detection using segmented ground truth and an attention mechanism.

Main Results:

• DRFA-Net demonstrated superior performance in WBC classification tasks.
• The model achieved higher accuracies compared to existing state-of-the-art methods on public datasets.
• Feature aggregation and discriminative region detection significantly boosted classification performance.

Conclusions:

• DRFA-Net effectively overcomes common challenges in WBC image analysis.
• The proposed method offers a robust and accurate solution for automated WBC classification.
• This approach holds promise for advancing pathological diagnosis through improved computational methods.