Background Multi-parametric magnetic resonance imaging may improve the detection of prostate cancer. The aim of this work is to compare between PI-RADS 3–5 and PI-RADS 4–5 as a threshold for targeted prostatic biopsy. Methods This is a prospective clinical study that included 40 biopsy-naïve patients referred for prostate biopsy. Patients underwent prebiopsy multi-parametric (mp-MRI), followed by 12-core transrectal ultrasound-guided systematic biopsy and cognitive MRI/TRUS fusion targeted biopsy from each detected lesion. The primary endpoint was to assess the diagnostic accuracy of the PI-RAD 3–4 versus PI-RADS 4–5 lesion by mpMRI for prostate cancer detection in biopsy-naive men. Results The overall prostate cancer detection rate and the clinically significant cancer detection rate were 42.5% and 35%, respectively. Targeted biopsies from PI-RADS 3–5 lesions showed a sensitivity of 100%, specificity of 44%, positive predictive value of 51.7%, and negative predictive value of 100%. Restricting targeted biopsies to PI-RADS 4–5 lesions resulted in a decrease in sensitivity and negative predictive value to 73.3% and 86.2%, respectively, while specificity and positive predictive value were increased to 100% for both parameters which was statistically significant (P value < 0.0001 and P value = 0.004, respectively). Conclusions Limiting the TBs to PI-RADS 4–5 lesions improves the performance of mp-MRI in the detection of prostate cancer especially aggressive tumors.