A deep learning approach for ovarian cancer detection and classification based on fuzzy deep learning

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Ovarian cancer detection and classification rely on individual oncologists' interpretations of histopathological whole slide images.
• A standardized and accurate automated system is crucial for early ovarian cancer detection and improved decision-making.

Purpose of the Study:

• To develop an automated system for accurate and standardized detection and classification of ovarian cancer using histopathological images.
• To enhance early diagnosis capabilities through advanced computational methods.

Main Methods:

• Feature extraction from histopathology images using the ResNet-50 model.
• Recursive feature elimination with a decision tree to refine extracted features.
• Integration of deep learning with fuzzy logic for final image classification.
• Optimization of network weights using Adam optimizers.

Main Results:

• The system achieved high diagnostic performance with 98.99% accuracy, 99% sensitivity, 98.96% specificity, and 99% F1-score.
• The dataset comprised 288 hematoxylin and eosin (H&E) stained whole slide images from 78 patients.
• The model demonstrated potential in classifying effective (162) and invalid (126) WSIs.

Conclusions:

• The proposed fuzzy deep-learning classifier shows significant promise for the accurate prediction of ovarian cancer.
• Automated systems can provide a more standardized and accurate approach to histopathological analysis in oncology.
• This technology has the potential to aid oncologists in early ovarian cancer detection and classification.