A Multi-Stage Hybrid Learning Model with Advanced Feature Fusion for Enhanced Prostate Cancer Classification.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Oncology

Background:

• Prostate cancer (PCa) is a leading cause of cancer death in men, presenting diagnostic challenges due to imaging variability.
• Magnetic Resonance Imaging (MRI) is crucial for PCa detection, but accurate classification requires integrating diverse feature types.
• Combining deep learning features (CNNs) with handcrafted descriptors (HOG) is vital for enhanced computer-aided diagnosis.

Purpose of the Study:

• To develop a multi-stage hybrid learning model for improved PCa diagnosis using MRI.
• To investigate feature reduction and classification techniques for optimal diagnostic performance.
• To enhance the accuracy and reliability of computer-aided diagnosis for prostate cancer.

Main Methods:

• Integrated deep features from CNNs with handcrafted texture descriptors (e.g., HOG).
• Employed dimensionality reduction techniques like Singular Value Decomposition (SVD) on the fused feature space.
• Benchmarked various machine learning classifiers and validated the framework using k-fold cross-validation.

Main Results:

• The hybrid model significantly outperformed individual deep or handcrafted feature approaches.
• Achieved exceptional performance metrics: 99.74% accuracy, 99.87% specificity, 99.87% precision, 99.61% sensitivity, and 99.74% F1-score on the TPP dataset.
• Demonstrated superior diagnostic capabilities for binary classification tasks.

Conclusions:

• The proposed hybrid model offers a robust and generalizable solution for PCa diagnosis.
• Effective integration of complementary features, dimensionality reduction, and optimized classification enhances diagnostic accuracy.
• The model shows strong potential for integration into clinical decision-support systems.