Deep fine-KNN classification of ovarian cancer subtypes using efficientNet-B0 extracted features: a comprehensive...

Area of Science:

• Oncology
• Computational Biology
• Medical Imaging

Background:

• Accurate classification of ovarian cancer subtypes is crucial for effective treatment and patient outcomes.
• Histopathological image analysis is a cornerstone of cancer diagnosis, but subtype differentiation can be challenging.

Purpose of the Study:

• To develop and validate a robust deep learning model for precise classification of five distinct ovarian cancer subtypes.
• To integrate EfficientNet-B0 deep features with a fine-tuned k-nearest neighbor (KNN) classifier for enhanced diagnostic accuracy.

Main Methods:

• Utilized the UBC-OCEAN dataset comprising 725 histopathological images of five ovarian cancer subtypes.
• Employed EfficientNet-B0 for deep feature extraction and a fine-KNN approach for classification.
• Split the dataset into 80% for training and 20% for testing, evaluating performance using accuracy, AUC, and LR+.

Main Results:

• Achieved 100% accuracy in both validation and testing phases.
• Demonstrated high Area Under the Curve (AUC) values across subtypes (e.g., 0.94 for MC).
• Reported significant Positive Likelihood Ratios (LR+) indicating strong diagnostic utility for subtypes like CC (27.294).

Conclusions:

• The integrated deep learning and KNN model effectively classifies ovarian cancer subtypes with exceptional accuracy.
• The model shows promise as a valuable tool for aiding in the diagnosis of ovarian cancer.
• Findings support the advancement of precision medicine through improved diagnostic capabilities in ovarian cancer research.