Diagnosis data of MulmoU-Net models for prostate cancer detection and segmentation on multi-sequence MRI

We have developed MulmoU-Net models for prostate cancer detection and segmentation on T2-weighted imaging alone, T2-weighted imaging plus dynamic contrast-enhanced imaging, biparametric MRI, and multiparametric MRI. Raw data is presented in this publication. Clinical data was collected from patients who were on suspicion of clinically localized prostate cancer and underwent multiparametric MRI followed by MRI-ultrasound fusion targeted biopsy and systematic 12-core biopsy. Region-based sensitivity and positive predictive value of each model was analyzed using data of 664 radiologist-identified regions with PIRADS scores ≥3. Regarding the specificity and the negative predictive value, whether a region was free from disease or not was examined in each prostatic side.