A New Framework for Precise Identification of Prostatic Adenocarcinoma

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Prostate cancer is a leading cause of cancer death globally.
• Early detection is crucial for patient survival.
• Accurate differentiation between malignant and benign prostate conditions is challenging.

Purpose of the Study:

• To develop a comprehensive, noninvasive computer-aided diagnosis (CAD) framework for precise differentiation of prostate cancer.
• To integrate multi-modal MRI features with clinical data for enhanced diagnostic accuracy.

Main Methods:

• Developed a CAD system integrating diffusion-weighted (DW) and T2-weighted (T2W) MRI modalities.
• Combined apparent diffusion coefficient (ADC) maps, texture features, and shape features (spherical harmonics).
• Integrated features with Prostate-Specific Antigen (PSA) screening results and evaluated using machine learning classifiers (SVM, RF, DT, LDA) on 80 patients.

Main Results:

• The Support Vector Machine (SVM) model with a combined, feature-selected set achieved 88.75% accuracy, 81.08% sensitivity, and 95.35% specificity.
• The integrated system demonstrated superior diagnostic performance compared to individual feature sets and other classifiers.
• Cross-validation (leave-one-out, 10-fold, 5-fold) confirmed the system's robustness and generalizability.

Conclusions:

• The developed CAD framework offers a reliable and accurate noninvasive method for diagnosing prostate cancer.
• Integration of multi-modal MRI data and clinical information significantly improves diagnostic capabilities.
• The system shows potential for improving early detection and treatment of prostatic adenocarcinoma.