A modified deep learning approach for seminal vesicle region localization in prostate MRI

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Male Reproductive Health

Background:

• Seminal vesicle evaluation is vital for diagnosing male infertility and carcinoma.
• Manual Magnetic Resonance Imaging (MRI) analysis is time-consuming and prone to interobserver variability.
• Automated methods are needed to enhance efficiency and consistency in seminal vesicle assessment.

Purpose of the Study:

• To develop and evaluate a deep learning model for automated localization of the seminal vesicle region in prostate MRI.
• To establish a robust region-level localization method as a foundation for detailed analysis.
• To address the gap in deep learning-based seminal vesicle localization from MRI.

Main Methods:

• A modified ResNet-34 deep learning model was employed for automated seminal vesicle localization.
• A sliding window approach was used to detect high-confidence regions.
• Performance was assessed using classification accuracy, inference time, and true positive (TP) coverage.

Main Results:

• The modified ResNet-34 model achieved high TP coverage (0.885) and classification accuracy (0.979).
• The model demonstrated improved localization with minimal computational overhead.
• Heatmap visualizations confirmed the model's focus on relevant anatomical structures.

Conclusions:

• The proposed deep learning approach offers a practical solution for reducing manual effort and interobserver variability in seminal vesicle assessment.
• This method provides a reliable foundation for subsequent segmentation and abnormality detection in prostate MRI.
• Future research may involve 3D CNNs or multi-view imaging for further performance enhancement.