Avoiding Unnecessary Systematic Biopsy in Clinically Significant Prostate Cancer: Comparison Between MRI‐Based Radiomics Model and PI‐RADS Category

• Published in: Journal of magnetic resonance imaging Vol. 57; no. 2; pp. 578 - 586
• Main Authors: Cheng, Xueqing, Chen, Yuntian, Xu, Hui, Ye, Lei, Tong, Shun, Li, Haixia, Zhang, Tianjing, Tian, Song, Qi, Jin, Zeng, Hao, Yao, Jin, Song, Bin
• Format: Journal Article
• Language: English
• Published: Hoboken, USA John Wiley & Sons, Inc 01.02.2023; Wiley Subscription Services, Inc
• Subjects: Aged; Biopsy; Calibration; Clinical significance; Cognitive ability; Decision analysis; Diffusion coefficient; Field strength; Humans; Image-Guided Biopsy - methods; Lesions; Magnetic resonance imaging; Magnetic Resonance Imaging - methods; Male; Middle Aged; MRI‐targeted biopsy; PI‐RADS; Population studies; Prostate; Prostate - diagnostic imaging; Prostate - pathology; Prostate cancer; Prostatic Neoplasms - diagnostic imaging; Prostatic Neoplasms - pathology; Radiomics; Retrospective Studies; Statistical analysis; Statistical tests; systematic biopsy; Training; radiomics; prostate cancer; systematic biopsy; PI-RADS; MRI-targeted biopsy
• ISSN: 1053-1807; 1522-2586
• DOI: 10.1002/jmri.28333

Background MRI‐targeted biopsy (MRTB) improves the clinically significant prostate cancer (csPCa) detection rate with fewer biopsy cores in men with suspected PCa. However, whether concurrent systematic biopsy (SB) can be avoided in patients undergoing MRTB remains unclear. Purpose To evaluate the potential value of MRI‐based radiomics models in avoiding unnecessary SB in biopsy‐naïve patients. Study type Retrospective. Population A total of 226 patients (mean age 66.6 ± 9.02 years) with suspicion of PCa (PI‐RADS score ≥ 3) and received combined cognitive MRTB with SB were retrospectively recruited and randomly divided into training (N = 180) and test (N = 46) cohorts at an 8:2 ratio. Field Strength/Sequence A 3.0 T, biparametric MRI (bpMRI) including T2‐weighted imaging (T2WI) and apparent diffusion coefficient (ADC) map. Assessment The whole prostate gland (PG) and the index lesion (IL) were delineated. Three radiomics models of bpMRIPG, bpMRIIL, and bpMRIPG+IL were constructed, respectively, and the performance of each radiomics model was compared with that of PI‐RADS assessment. Statistical Tests The least absolute shrinkage and selection operator (LASSO) regression method was used to select texture features. The area under the curve (AUC) and decision curve analysis were used to estimate the models. Results The bpMRIPG+IL radiomics model exhibited good discrimination, calibration, and net benefits, which would reduce the SB biopsy in 71.2% and 71.4% of men with PI‐RADS ≥ 5 lesions in the training and test cohorts, respectively. Data Conclusion A bpMRIPG+IL radiomics model may outperform PI‐RADS category in help reducing unnecessary SB in biopsy‐naïve patients. Evidence Level 3 Technical Efficacy Stage 6