Multiparametric MRI-based prostate cancer classification using transfer learning, feature fusion, and ensemble...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Oncology
• Radiology

Background:

• Prostate cancer (PCa) is a leading cause of cancer mortality in men worldwide.
• Accurate classification of PCa into clinically significant (CS) and clinically insignificant (CiS) is crucial for treatment decisions.
• Multiparametric magnetic resonance imaging (mpMRI) is vital for detecting and characterizing PCa lesions.

Purpose of the Study:

• To propose an effective framework for PCa classification using mpMRI.
• To differentiate between CS and CiS PCa lesions for improved patient management.
• To leverage transfer learning and feature fusion for enhanced diagnostic accuracy.

Main Methods:

• Utilized transfer learning with pre-trained VGG19 and Vision Transformer (ViT) models for feature extraction from nine mpMRI sequences (PROSTATEx dataset).
• Employed feature fusion and Support Vector Machine (SVM) classification.
• Implemented an ensemble approach combining multiple model outputs for final classification.

Main Results:

• The proposed framework achieved a promising Area Under the Curve (AUC) score of 0.85 with the ensemble output.
• Incorporating higher b-values in Diffusion-Weighted Imaging (DWI) sequences was found to be important.
• Combining CNN-based (VGG19) and transformer-based (ViT) features proved effective.

Conclusions:

• The developed framework shows potential for supporting radiologists in accurate and efficient PCa diagnosis.
• The integration of VGG19 and ViT features enhances the classification of CS and CiS lesions.
• This approach can aid in personalized treatment strategies for prostate cancer patients.