BackgroundCognitive fusion MRI-guided targeted biopsy (cTB) has been widely used in the diagnosis of prostate cancer (PCa). However, cTB relies heavily on the operator’s experience and confidence in MRI readings. Our objective was to compare the cancer detection rates of MRI artificial intelligence-guided cTB (AI-cTB) and routine cTB and explore the added value of using AI for the guidance of cTB.MethodsThis was a prospective, single-institution randomized controlled trial (RCT) comparing clinically significant PCa (csPCa) and PCa detection rates between AI-cTB and cTB. A total of 380 eligible patients were randomized to the AI-cTB group (n = 191) or the cTB group (n = 189). The AI-cTB group underwent AI-cTB plus systematic biopsy (SB) and the cTB group underwent routine cTB plus SB. The primary outcome was the detection rate of csPCa. The reference standard was the pathological results of the combination of TB (AI-cTB/cTB) and SB. Comparisons of detection rates of csPCa and PCa between groups were performed using the chi-square test or Fisher’s exact test.ResultsThe overall csPCa and PCa detection rates of the whole inclusion cohort were 58.8% and 61.3%, respectively. The csPCa detection rates of TB combined with SB in the AI-cTB group were significantly greater than those in the cTB group at both the patient level (58.64% vs. 46.56%, p = 0.018) and per-lesion level (61.47% vs. 47.79%, p = 0.004). Compared with cTB, the AI-cTB could detect a greater proportion of patients with csPCa at both the per-patient level (69.39% vs. 49.71%, p < 0.001) and per-lesion level (68.97% vs. 48.57%, p < 0.001). Multivariate logistic analysis indicated that compared with the cTB, the AI-cTB significantly improved the possibility of detecting csPCa (p < 0.001).ConclusionsAI-cTB effectively improved the csPCa detection rate. This study successfully integrated AI with TB in the routine clinical workflow and provided a research paradigm for prospective AI-integrated clinical studies.Trial registrationClinicalTrials.gov, NCT06362291.
Read full abstract