Comprehensive Evaluation of MRI-AI in Prostate Cancer Diagnosis

In recent years, there have been remarkable advancements in the field of artificial intelligence (AI) techniques, particularly in the medical domain. These AI techniques have demonstrated the ability to significantly enhance various medical tasks, such as tumor detection, classification, and prognosis prediction. Increasing evidence supports the ability of AI to facilitate precise diagnosis of PCa and assist in therapeutic decisions. Compared with doctors, AI has the potential to identify not only holistic tumor morphology but also task-specific and granular radiological patterns that cannot be detected by the naked eye. Therefore, AI has great potential to reduce inconsistencies between observers and improve diagnostic accuracy. Previous AI studies at our institution have developed deep learning-based AI models trained on MR images that achieve good performance in the detection and localization of clinically significant prostate cancer (csPCa). Furthermore, the trained AI algorithms were embedded into proprietary structured reporting software, and radiologists simulated their real-life work scenarios to interpret and report the PI-RADS category of each patient using this AI-based software. However, the data is mostly retrospective. The capability of detecting the suspicious lesions on MRI, guiding the prostate targeted biopsy, and optimizing the biopsy scheme warrants further investigation.

The goal of this real-world prospective diagnostic study is to comprehensively evaluate the value of MRI artificial intelligence (MRI-AI) in assisting the diagnosis of prostate cancer (PCa). The main questions it aims to answer are:

Does MRI-AI promote the accurate diagnosis and treatment of prostate cancer? What's the capability of prostate MRI-AI in calculating the prostate volumn? What's the value of prostate MRI-AI assistant diagnosis system in detecting the suspicious lesions on MRI and guiding prostate targeted biopsy? What's the value of prostate MRI-AI assistant diagnosis system in predicting the pathological results of prostate targeted biopsy? Researchers will compare the cancer detection rates of suspicious lesions detected by MRI-AI and senior radiologists.

Participants will:

Receive combination of systematic biopsy and targeted biopsy.

Inclusion criteria

• The age of the patient is between 45 and 85.
• Patients with complete magnetic resonance imaging (MRI) data, qualified image quality control.
• Patients were in accordance with the indication of prostate biopsy, including patients with suspicious prostate nodes found by digital rectal examination (DRE), the suspicious lesions found by transrectal ultrasound (TRUS) or MRI, total prostate-specific antigen (tPSA) >10ng/mL, tPSA 4-10ng/mL with free-to-total PSA ratio (f/tPSA) <0.16 or PSA density (PSAD) >0.15.
• Patients had no history of prior prostate surgery or biopsy.
• The PSA of patients should be ≤20 ng/mL.
• The prostate biopsy pathological results of above lesions were complete. The time interval between targeted prostate biopsy and prostate MRI examination should not exceed one month.
• Patients with complete clinical information.

Exclusion criteria

• The clinicopathological information and MRI data was unqualified or incomplete.
• Patients had received radiotherapy, chemotherapy, androgen deprivation therapy, or surgery treatment before prostate MRI examination or prostate biopsy.
• Patients received prior prostate biopsy.
• Patients had contraindications to MRI or prostate biopsy.
• Patients were not in accordance with the indication of prostate biopsy.