Multimodal approach to optimize biopsy decision-making for PI-RADS 3 lesions on multiparametric MRI

Area of Science:

• Radiology
• Artificial Intelligence
• Oncology

Background:

• Prostate cancer diagnosis often relies on Prostate Imaging Reporting and Data System (PI-RADS) scores.
• PI-RADS 3 lesions present diagnostic uncertainty, leading to decisions regarding prostate biopsy.
• Accurate identification of clinically significant prostate cancer (csPCa) is crucial to avoid underdiagnosis or overtreatment.

Purpose of the Study:

• To develop and evaluate a multimodal approach for determining the necessity of prostate biopsy in patients with PI-RADS 3 lesions.
• This approach integrates clinical parameters with a biparametric MRI-based artificial intelligence (AI) model.
• The goal is to improve the accuracy of csPCa detection and optimize biopsy decisions.

Main Methods:

• A retrospective study of a prospectively recruited patient cohort with PI-RADS 3 lesions was conducted.
• Data included demographic information, PSA levels, PSA density (PSAD), MRI interpretations, and AI evaluations.
• The AI model's performance in detecting csPCa was compared against MRI/US fusion-guided biopsy results, using statistical methods like Fisher's exact test and logistic regression.

Main Results:

• The study included 248 patients with 312 PI-RADS 3 lesions; 44 were csPCa.
• The AI model demonstrated a negative predictive value (NPV) of 89.2% for csPCa across all lesions.
• Combining AI with PSAD (≥0.15 ng/mL²) significantly improved sensitivity to 77.8% while maintaining a high NPV of 93.1%.

Conclusions:

• A multimodal approach combining AI and PSAD shows promise in enhancing prostate biopsy decision-making for PI-RADS 3 lesions.
• This strategy can help minimize the occurrence of missed csPCa.
• It also has the potential to reduce the number of unnecessary prostate biopsies performed.