Objective:

To design and evaluate ensemble deep learning models for the automatic classification of pleural cytology images into malignant and negative for malignancy categories, addressing the critical need for improved diagnostic accuracy.

Approach:

Key Findings:

• The ResNet + DenseNet ensemble with soft voting and 300% data augmentation achieved the highest accuracy (96.2% on the local dataset; 89.6% on the external dataset), indicating the effectiveness of ensemble methods in this context.
• The ensemble approach outperformed individual models across all scenarios, highlighting the advantages of combining multiple architectures.
• Data augmentation significantly improved generalization and robustness, suggesting its critical role in enhancing model performance.

Interpretation:

The proposed ensemble-based approach supports cytological diagnosis, potentially reducing diagnostic uncertainty in pleural carcinoma detection and improving patient outcomes.

Limitations:

• The study is limited to specific datasets which may affect generalizability; future research should explore diverse datasets.
• The reliance on pre-trained models may not capture all domain-specific features, indicating a need for further model training on specialized datasets.

Conclusion:

Ensemble deep learning models, optimized with data augmentation, can provide accurate and reproducible diagnostic support for pleural cytology, with significant potential for deployment in low-resource healthcare settings.

Attribution Notice

This content is an AI-generated, fully rewritten summary based on a published scholarly article. It does not reproduce the original text and is not a substitute for the original publication. Readers are encouraged to consult the source for full context, data, and methodology.