Lesion-guided selective multi-modal integration for prostate cancer segmentation and PI-RADS grading in MP-MRI

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Oncology

Background:

• Prostate cancer (PCa) is a major global health concern.
• Accurate segmentation and grading of PCa lesions in multiparametric MRI (mp-MRI) are critical for diagnosis and treatment.

Purpose of the Study:

• To develop and validate an automated model for PCa lesion segmentation and Prostate Imaging Reporting and Data System (PI-RADS) grading in mp-MRI.

Main Methods:

• Proposed a Lesion-guided Selective Multi-modal Integration (LeSMI) module with Dynamic Modality Weighting (DMW) and Localized Lesion Attention (LLA).
• Employed a two-stage Prostate Cancer Segmentation and Grading (PCaSG) strategy.
• Validated on Prostate158 and PI-CAI Challenge datasets, including 5-fold cross-validation.

Main Results:

• Achieved Dice Similarity Coefficient (DSC) of 51.30% and quadratic-weighted kappa (QWK) of 62.48% on Prostate158.
• Achieved DSC of 43.81% and QWK of 42.98% on PI-CAI.
• Demonstrated improvements over state-of-the-art models, with up to 2% DSC and 17% QWK gains on Prostate158, and 4% DSC and 3% QWK gains on PI-CAI.

Conclusions:

• The model shows robustness in handling diverse lesion presentations and reliable assessments, indicating significant clinical applicability.
• Offers advancements in segmentation accuracy and PI-RADS grading, addressing inter-reader variability and expertise requirements.
• Promises enhanced early detection, accurate risk assessment, and improved patient outcomes in prostate cancer management.