Magnetic Resonance Imaging-based Prostate Cancer Diagnosis Using Principal Component Analysis and Machine Learning...

Area of Science:

• Medical Imaging
• Machine Learning
• Oncology

Background:

• Prostate cancer poses a significant threat to men's health, underscoring the need for precise diagnostic tools.
• Accurate differentiation between benign and malignant prostate lesions is crucial for effective treatment planning.

Purpose of the Study:

• To assess the efficacy of principal component analysis (PCA) integrated with machine learning (ML) for classifying prostate lesions using MRI.
• To compare the performance of PCA-transformed features against raw pixel and radiomics features for lesion classification.

Main Methods:

• Magnetic resonance imaging (MRI) data from 26 prostate cancer patients were analyzed.
• Principal component analysis (PCA) was employed for feature extraction from MRI images.
• Random Forest, SVM, and KNN classifiers were utilized to categorize lesions based on extracted features.

Main Results:

• PCA-based features showed significant separation between benign and malignant tissues (P < 2.2 × 10⁻¹⁶).
• The Random Forest classifier achieved the highest diagnostic performance with an AUC of 0.909.
• PCA-enhanced models outperformed raw pixel-based methods, with radiomics features providing further improvements.

Conclusions:

• PCA-based feature extraction enhances classification accuracy by reducing noise and dimensionality in MRI data.
• The proposed PCA-ML framework provides an interpretable and efficient method for prostate cancer diagnosis.
• This approach offers a clinically valuable tool for early and accurate detection of prostate cancer.