Automated prostate cancer localization without the need for peripheral zone extraction using multiparametric MRI

Area of Science:

• Radiology and Medical Imaging
• Oncology
• Computer-Aided Diagnosis

Background:

• Multiparametric MRI (mpMRI) offers superior prostate cancer localization compared to TRUS.
• Current automated segmentation methods require manual extraction of the prostate's peripheral zone (PZ).
• Manual PZ extraction is labor-intensive and time-consuming, hindering efficient automated analysis.

Purpose of the Study:

• To develop an automated method for prostate cancer segmentation that eliminates the need for manual PZ extraction.
• To integrate spatial and geometric tumor information with mpMRI data for improved localization accuracy.
• To leverage T2-weighted MRI, DWI, and DCE-MRI for comprehensive prostate cancer assessment.

Main Methods:

• A novel 'location map' feature was introduced, encoding spatial information through nonlinear transformation of pixel coordinates.
• This location map was fused with mpMRI data (T2w, DWI, DCE-MRI) for automated tumor localization.
• The algorithm was validated on mpMRI data from 20 patients with biopsy-confirmed prostate cancer.

Main Results:

• The proposed method achieved a specificity of 0.84 and a sensitivity of 0.80 for prostate cancer detection.
• A Dice coefficient of 0.42 was obtained, indicating successful tumor segmentation.
• Performance was comparable or superior to methods requiring manual PZ segmentation.

Conclusions:

• Fusing spatial information enables accurate tumor outlining without manual PZ extraction.
• The developed method demonstrates significant success in automated prostate cancer localization.
• Experimental results confirm the effectiveness and efficiency of the proposed approach for clinical application.