PSP net-based automatic segmentation network model for prostate magnetic resonance imaging.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Oncology

Background:

• Prostate cancer diagnosis relies heavily on accurate interpretation of medical images.
• Improving the precision of Magnetic Resonance Imaging (MRI) segmentation is crucial for early detection and treatment planning.
• Current segmentation methods may have limitations in accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a novel prostate MRI segmentation model using Pyramid Scene Parsing Network (PSP Net).
• To enhance the accuracy of automatic segmentation for prostate cancer detection.
• To compare the proposed PSP Net model against established segmentation networks like FCN and U-Net.

Main Methods:

• Utilized a dataset of 270 prostate MRI images.
• Applied Contrast Limited Adaptive Histogram Equalization (CLAHE) for image enhancement.
• Implemented and compared a PSP Net-based segmentation model with Fully Convolutional Networks (FCN) and U-Net, using metrics such as segmentation accuracy, under-segmentation rate, over-segmentation rate, and Receiver Operating Characteristic (ROC) curves.

Main Results:

• The PSP Net model achieved the highest segmentation accuracy (0.9865).
• PSP Net demonstrated superior performance with lower over-segmentation (0.0023) and under-segmentation (0.1111) rates compared to FCN and U-Net.
• The ROC curve analysis showed PSP Net's Area Under the Curve (AUC) was 0.9427, indicating better diagnostic capability.

Conclusions:

• The PSP Net-based prostate MRI automatic segmentation model significantly improves segmentation accuracy.
• This advanced model has the potential to reduce the workload for radiologists.
• The findings support the further clinical promotion and adoption of the PSP Net model for prostate cancer diagnosis.