An efficient Fusion-Purification Network for Cervical pap-smear image classification.

Area of Science:

• Medical imaging analysis
• Computational pathology
• Artificial intelligence in diagnostics

Background:

• Deep learning models show promise in cervical cell image analysis.
• Challenges remain in discriminating cervical cells due to variations.
• Current methods often overlook global morphological information.

Purpose of the Study:

• To develop a novel cervical cell classification model.
• To improve feature discrimination by integrating texture and morphology.
• To introduce the Cervical Cytopathology Image Dataset (CCID).

Main Methods:

• Propose a model focusing on purified fusion information.
• Integrate detailed texture and morphological structure features (cervical pathology information fusion).
• Design a cervical purification bottleneck module to address data redundancy and bias.

Main Results:

• The proposed model achieves superior performance compared to state-of-the-art methods.
• Experiments conducted on two real-world datasets validate the model's effectiveness.

Conclusions:

• The developed method aids pathologists in accurate cervical smear evaluation.
• The model effectively captures localized informative differences and represents discriminative features.