PI-RADS 4 and 5: may radiomics help improve radiologists’ performance in detecting clinically significant cancers?
Number:28ESURABS0055
Type:Scientific Electronic Presentation
Authors: Paolo Niccolò Franco, Pietro Andrea Bonaffini, Andrea Corsi, Elisabetta de Bernardi, Davide Ippolito, Sandro Sironi
Keywords:PI-RADS v.2.1, PI-RADS 4 and 5 lesions, Multiparametric prostate MRI, Radiomics.
SECTIONS
Objectives

Prostate cancer (PCa) is the most frequently diagnosed cancer among adult men, and its incidence is increasing worldwide. The risk of PCa is related to age, family history, ethnicity, prostate-specific antigen (PSA) levels, and findings on digital rectal examination.

Multiparametric Magnetic Resonance Imaging (mp-MRI) is considered the imaging gold standard for PCa diagnosis, staging, and follow-up during active surveillance. The Prostate Imaging Reporting and Data System (PI-RADS) version 2.1 represents the cornerstone of prostate MRI reading. PI-RADS criteria require the interpreting radiologist to score prostate lesions into five categories that represent their probability of being clinically significant cancers (csPCa), defined as tumours showing a Gleason score (GS) equal to or greater than 7 at histopathological analysis....

...read more

Material and methods

This study included consecutive patients who underwent a pelvic mp-MRI performed at our institution for suspected PCa or during active surveillance between June 2016 and March 2021. The inclusion criteria were as follows: (1) a lesion scored as PI-RADS 4 or 5 according to PI-RADS v2.1 guidelines; (2) mp-MRI scan performed before radical prostatectomy or radiotherapy; (3) available histopathological data from MRI-targeted biopsies or prostatectomy specimens. Exclusion criteria were: (1) patients with absolute contraindications to MRI (i.e., pacemakers/defibrillator carriers); (2) poor image quality on the T2-weighted (T2W) and/or the diffusion-weighted imaging (DWI) sequences.

Data regarding patients' demographic, laboratory, pathological, and conventional MRI parameters were collected: patients' age, the most recent serological value of PSA (ng/mL), PSA...

...read more

Fig. 1 ROI delineation and the generated 3D VOI in a peripheral zone lesion (white arrows). In the axial T2-wieighted image (a), a markedly hypointense lesion exceeding 15 mm in length is observed in anterior/posterior-lateral right peripheral zone of prostate midportion. The lesion showed a high signal intensity on DWI (c) and a low signal intensity in the ADC map (d). The lesion was scored as PI-RADS 5 and manually contoured on T2-weighted, and diffusion-weighted images (b, e) and a 3D VOI of the whole lesion (f) was generated. ROI: region of interest; 3D VOI: three-dimensional volume of interest; DWI: diffusion-weighted imaging; ADC: apparent diffusion coefficient; PI-RADS: Prostate Imaging Reporting and Data System.
Fig. 2 ROI delineation and the generated 3D VOI in a transitional zone lesion. In the right transitional zone of the gland apex, an ill-defined with a major diameter >15 mm is observed (white arrows). The lesion shows a moderately low signal on T2W sequence (a), high signal on DWI (c), and signal loss on the ADC map (d), being consistent with a PI-RADS 5 category. It was segmentated on T2W and DWI (b, e), and a 3D VOI of the lesion (f) was constructed. ROI: region of interest; 3D VOI: three-dimensional volume of interest; T2W: T2-weighted; DWI: diffusion-weighted imaging; ADC: apparent diffusion coefficient; PI-RADS: Prostate Imaging Reporting and Data System.
Results

Overall, 945 consecutive patients underwent a prostate mp-MRI for suspected PCa or active surveillance during the study period. In this cohort, 112 patients were diagnosed with at least one PI-RADS 4 or 5 lesion. Thirteen patients were excluded due to T2W sequence (n=2) and ADC maps' poor image quality (n=5) or for lack of histopathological data (n=6). Therefore, the final cohort of the study included 99 patients (median age 69 years, IQR 59-79). Among them were diagnosed: one PI-RADS 4 or 5 lesion in 89 patients; two lesions in 8 patients; three lesions in 2 patients. Thus, 111 lesions (1.12 lesion for patient) were finally scored as PI-RADS 4 or 5. Seventy-four lesions (66.7%) were PI-RADS 4 (68 [91.8%] in the PZ and 6 [8.1%] in the TZ), while 37 (33.3%) were PI-RADS 5 (18 [48.7%] in the PZ and 19 [51.3%] in the TZ). At histopathological analysis, 79 lesions (71%) resulted...

...read more

Conclusions

MRI radiomic features based on axial T2-weighted and diffusion-weighted imaging show potential in supporting radiologists' subjective detection of clinically significant cancers among lesions scored as PI-RADS 4 and 5 lesions based on PI-RADS v2.1 criteria. Features identified as correlated with clinically significant cancers in the prostate zone were different from those identified for transitional zone lesions. Tests on wider samples are needed to validate this study's results.