mEPE-score: a comprehensive grading system for predicting pathologic extraprostatic extension of prostate cancer at multiparametric magnetic resonance imaging.

OBJECTIVE: To investigate the diagnostic accuracy of the PI-RADS v2.1 multiparametric magnetic resonance imaging (mpMRI) features in predicting extraprostatic extension (mEPE) of prostate cancer (PCa), as well as to develop and validate a comprehensive mpMRI-derived score (mEPE-score). METHODS: We retrospectively reviewed all consecutive patients admitted to two institutions for radical prostatectomy for PCa with available records of mpMRI performed between January 2015 and December 2020. Data from one institution was used for investigating diagnostic performance of each mEPE feature using radical prostatectomy specimens as benchmark. The results were implemented in a mEPE-score as follows: no mEPE features: 1; capsular abutment: 2; irregular or spiculated margin: 3; bulging prostatic contour, or asymmetry of the neurovascular bundles, or tumor-capsule interface > 1.0 cm: 4; ≥ 2 of the previous three parameters or measurable extraprostatic disease: 5. The performance of mEPE features was evaluated using the five diagnostic parameters and ROC curve analysis. RESULTS: Two-hundred patients were enrolled at site 1 and 76 at site 2. mEPE features had poor sensitivities ranging from 0.08 (0.00-0.15) to 0.71 (0.59-0.83), whereas specificity ranged from 0.68 (0.58-0.79) to 1.00. mEPE-score showed excellent discriminating ability (AUC > 0.8) and sensitivity = 0.82 and specificity = 0.77 with a threshold of 3. mEPE-score had AUC comparable to ESUR-score (p = 0.59 internal validation; p = 0.82 external validation), higher than or comparable to mEPE-grade (p = 0.04 internal validation; p = 0.58 external validation), and higher than early-and-late-EPE (p < 0.0001 internal and external validation). There were no significant differences between readers having different expertise with EPE-score (p = 0.32) or mEPE-grade (p = 0.45), but there were significant differences for ESUR-score (p = 0.02) and early-versus-late-EPE (p = 0.03). CONCLUSIONS: The individual mEPE features have low sensitivity and high specificity. The use of mEPE-score allows for consistent and reliable assessment for pathologic EPE. KEY POINTS: • Individual PI-RADS v2.1 mpMRI features had poor sensitivities ranging from 0.08 (0.00-0.15) to 0.71 (0.59-0.83), whereas Sp ranged from 0.68 (0.58-0.79) to 1.00. • mEPE-score is an all-inclusive score for the assessment of pEPE with excellent discriminating ability (i.e., AUC > 0.8) and Se = 0.82, Sp = 0.77, PPV = 0.74, and NPV = 0.84 with a threshold of 3. • The diagnostic performance of the expert reader and beginner reader with pEPE-score was comparable (p = 0.32).

MRI index lesion radiomics and machine learning for detection of extraprostatic extension of disease: a multicenter study.

OBJECTIVES: To build a machine learning (ML) model to detect extraprostatic extension (EPE) of prostate cancer (PCa), based on radiomics features extracted from prostate MRI index lesions. METHODS: Consecutive MRI exams of patients undergoing radical prostatectomy for PCa were retrospectively collected from three institutions. Axial T2-weighted and apparent diffusion coefficient map images were annotated to obtain index lesion volumes of interest for radiomics feature extraction. Data from one institution was used for training, feature selection (using reproducibility, variance and pairwise correlation analyses, and a correlation-based subset evaluator), and tuning of a support vector machine (SVM) algorithm, with stratified 10-fold cross-validation. The model was tested on the two remaining institutions' data and compared with a baseline reference and expert radiologist assessment of EPE. RESULTS: In total, 193 patients were included. From an initial dataset of 2436 features, 2287 were excluded due to either poor stability, low variance, or high collinearity. Among the remaining, 14 features were used to train the ML model, which reached an overall accuracy of 83% in the training set. In the two external test sets, the SVM achieved an accuracy of 79% and 74% respectively, not statistically different from that of the radiologist (81-83%, p = 0.39-1) and outperforming the baseline reference (p = 0.001-0.02). CONCLUSIONS: A ML model solely based on radiomics features demonstrated high accuracy for EPE detection and good generalizability in a multicenter setting. Paired to qualitative EPE assessment, this approach could aid radiologists in this challenging task. KEY POINTS: • Predicting the presence of EPE in prostate cancer patients is a challenging task for radiologists. • A support vector machine algorithm achieved high diagnostic accuracy for EPE detection, with good generalizability when tested on multiple external datasets. • The performance of the algorithm was not significantly different from that of an experienced radiologist.

Whole body MRI with qualitative and quantitative analysis of DWI for assessment of bone marrow involvement in lymphoma.

AIM: Our study aimed to investigate the role of qualitative and quantitative whole body MRI with DWI for assessment of bone marrow involvement (BMI) in newly diagnosed lymphoma using FDG PET-CT and bone marrow biopsy (BMB) as reference standard. MATERIALS AND METHODS: We retrospectively evaluated 56 patients with newly diagnosed lymphoma (21 Hodgkin's lymphoma and 35 non-Hodgkin's lymphoma) who underwent random unilateral BMB, FDG PET-CT and Wb-MRI-DWI for initial staging. In a patient-based analysis, results of Wb-MRI-DWI were compared with FDG PET-CT and BMB. For quantitative analysis, mean ADC values of posterior iliac crest were correlated with BMI and bone marrow cellularity. RESULTS: WB-MR-DWI obtained excellent concordance with FDG PET-CT both in HL (k = 1.000; 95% CI 1.000-1.000) and in DLBCL (k = 1.000; 95% CI 1.000-1.000). In other NHL, WB-MRI-DWI obtained a good correlation with BMB (k = 0.611; 95% CI 0.295-0.927) while FDG PET-CT had poor concordance (k = 0.067; 95% CI 0.372-0.505). WB-MR-DWI has no false negative errors but 4 false positive results consisting in focal lesions consensually reported by FDG PET-CT and resolved after therapy. No significant correlation between ADC mean value and BMI was found (p = 0.0586). CONCLUSION: Our data suggest that Wb-MRI-DWI is a valid technique for BMI assessment in lymphoma patients, thanks to its excellent concordance with FDG PET-CT and good concordance with BMB (superior than FDG PET-CT). If further investigations will confirm our results on larger patient groups, it could become a useful tool in the clinical workup.

Digital Breast Tomosynthesis (DBT) to Characterize MRI-Detected Additional Lesions Unidentified at Targeted Ultrasound in Newly Diagnosed Breast Cancer Patients.

OBJECTIVES: Preoperative breast magnetic resonance (MR) often generates additional suspicious findings needing further investigations. Targeted breast ultrasound (US) is the standard tool to characterize MR additional lesions. The purpose of this study is to evaluate the potential role of digital breast tomosynthesis (DBT) to characterize MR detected additional findings, unidentified at targeted breast US. METHODS: This prospective study included women who a) had biopsy-proven, newly diagnosed breast cancers detected at conventional 2D mammography and/or US, referred to breast MR for tumour staging; and b) had DBT if additional MR findings were not detected at targeted ('second look') US. RESULTS: In 520 patients, MR identified 164 (in 114 women, 22%) additional enhancing lesions. Targeted US identified 114/164 (69.5%) of these, whereas 50/164 (30.5%) remained unidentified. DBT identified 32/50 of these cases, increasing the overall characterization of MR detected additional findings to 89.0% (146/164). Using DBT the identified lesions were significantly more likely to be malignant than benign MR-detected additional lesions (p = 0.04). CONCLUSIONS: DBT improves the characterization of additional MR findings not identified at targeted breast US in preoperative breast cancer staging. KEY POINTS: • Targeted US identified 114 of 164 (69.5%) additional enhancing lesions at preoperative breast MRI. • DBT identified a further 32 of the 50 lesions unidentified on targeted US. • DBT improved the characterization of additional MR findings for breast cancer staging.