Prostate zonal segmentation in 1.5T and 3T T2W MRI using a convolutional neural network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Prostate cancer (PCa) diagnosis and treatment planning rely on accurate segmentation of prostate gland zones using MRI.
• Current segmentation methods may face challenges in consistency and accuracy.

Purpose of the Study:

• To evaluate a U-net based 2D convolutional neural network (CNN) for segmenting the central gland (CG) and peripheral zone (PZ) of the prostate.
• To assess the performance of the CNN across different MRI scanners and prostate regions.

Main Methods:

• A dataset of 40 patients' MRI scans was used, with images preprocessed (cropping, resampling, normalization).
• A U-net CNN architecture was employed for zonal segmentation.
• Performance was quantified using Dice Similarity Coefficient (DSC) and Mean Absolute Distance (MAD) in a fivefold cross-validation.

Main Results:

• The CNN achieved an overall DSC of 0.794 for CG and 0.692 for PZ, with MADs of 3.349 and 2.993, respectively.
• Segmentation accuracy (DSC) was higher in the midgland region compared to the apex and base for both CG and PZ.
• No significant performance difference was observed between the two MRI scanners used.

Conclusions:

• The U-net based CNN demonstrates potential for accurate prostate zonal segmentation in MRI.
• The algorithm shows clinical promise for improving PCa diagnosis and image-guided therapies.
• Further validation with larger, multi-vendor datasets is recommended.