A Machine Learning-Based Unenhanced Radiomics Approach to Distinguishing Between Benign and Malignant Breast Lesions Using T2-Weighted and Diffusion-Weighted MRI.

BACKGROUND: Breast MRI has been recommended as supplemental screening tool to mammography and breast ultrasound of breast cancer by international guidelines, but its long examination time and use of contrast material remains concerning. PURPOSE: To develop an unenhanced radiomics model with using non-gadolinium based sequences for detecting breast cancer based on T2-weighted (T2W) and diffusion-weighted (DW) MRI. STUDY TYPE: Retrospective analysis followed by retrospective and prospective cohorts study. POPULATION: 1760 patients: Of these, 1293 for model construction (n = 775 for training and 518 for validation). The remaining patients for model testing in internal retrospective (n = 167), internal prospective (n = 188), and external retrospective (n = 112) cohorts. FIELD STRENGTH/SEQUENCE: 3.0T MR scanners from two institution. T2WI, DWI, and first contrast-enhanced T1-weighted sequence. ASSESSMENT: AUCs in distinguishing breast cancer were compared between combined model with gadolinium agent sequence and unenhanced model. Subsequently, the AUCs in testing cohorts of unenhanced model was compared with two radiologists' diagnosis for this research. Finally, patient subgroup analysis in testing cohorts was performed based on clinical subgroups and different types of malignancies. STATISTICAL TESTS: Mann-Whitney U test, Kruskal-Wallis H test, chi-square test, weighted kappa test, and DeLong's test. RESULTS: The unenhanced radiomics model performed best under Gaussian process (GP) classifiers (AUC: training, 0.893; validation, 0.848) compared to support vector machine (SVM) and logistic, showing favorable prediction in testing cohorts (AUCs, 0.818-0.840). The AUCs for the unenhanced radiomics model were not statistically different in five cohorts from those of the combined radiomics model (P, 0.317-0.816), as well as the two radiologists (P, 0.181-0.918). The unenhanced radiomics model was least successful in identifying ductal carcinoma in situ, whereas did not show statistical significance in other subgroups. DATA CONCLUSION: An unenhanced radiomics model based on T2WI and DWI has comparable diagnostic accuracy to the combined model using the gadolinium agent. LEVEL OF EVIDENCE: 4 TECHNICAL EFFICACY: Stage 2.

Predictive model containing PI-RADS v2 score for postoperative seminal vesicle invasion among prostate cancer patients.

BACKGROUND: Seminal vesicle invasion (SVI) is considered to be one of most adverse prognostic findings in prostate cancer, affecting the biochemical progression-free survival and disease-specific survival. Multiparametric magnetic resonance imaging (mpMRI) has shown excellent specificity in diagnosis of SVI, but with poor sensitivity. The aim of this study is to create a model that includes the Prostate Imaging Reporting and Data System version 2 (PI-RADS v2) score to predict postoperative SVI in patients without SVI on preoperative mpMRI. METHODS: A total of 262 prostate cancer patients without SVI on preoperative mpMRI who underwent radical prostatectomy (RP) at our institution from January 2012 to July 2019 were enrolled retrospectively. The prostate-specific antigen levels in all patients were <10 ng/mL. Univariate and multivariate logistic regression analyses were used to assess factors associated with SVI, including the PI-RADS v2 score. A regression coefficient-based model was built for predicting SVI. The receiver operating characteristic curve was used to assess the performance of the model. RESULTS: SVI was reported on the RP specimens in 30 patients (11.5%). The univariate and multivariate analyses revealed that biopsy Gleason grade group (GGG) and the PI-RADS v2 score were significant independent predictors of SVI (all P<0.05). The area under the curve of the model was 0.746 (P<0.001). The PI-RADS v2 score <4 and Gleason grade <8 yielded only a 1.8% incidence of SVI with a high negative predictive value of 98.2% (95% CI, 93.0-99.6%). CONCLUSIONS: The PI-RADS v2 score <4 in prostate cancer patients with prostate-specific antigen level <10 ng/mL is associated with a very low risk of SVI. A model based on biopsy Gleason grade and PI-RADS v2 score may help to predict SVI and serve as a tool for the urologists to make surgical plans.

Cutoff Values of Prostate Imaging Reporting and Data System Version 2.1 Score in Men With Prostate-specific Antigen Level 4 to 10 ng/mL: Importance of Lesion Location.

INTRODUCTION: Multiparametric magnetic resonance imaging (mpMRI) has been shown to have a good performance in predicting cancer among patients with a prostate-specific antigen (PSA) level of 4 to 10 ng/mL. However, lesion location on mpMRI has never been separately considered. PATIENTS AND METHODS: Patients with PSA level of 4 to 10 ng/mL were prospectively enrolled and underwent transrectal ultrasound-guided prostate biopsy. Patient information was collected, and logistic regression analysis was performed to determine the predictive factors of clinically significant prostate cancer (csPCa). Patients were grouped by lesion location to determine the Prostate Imaging Reporting and Data System (PI-RADS) v2.1 cutoff value in predicting csPCa. RESULTS: Among 222 patients, 121 were diagnosed with PCa and 92 had csPCa. Age, prostate volume, PSA density, location (peripheral zone, csPCa only), and PI-RADS v2.1 score were correlated with PCa and csPCa, and PI-RADS v2.1 score was the best predictor. A PI-RADS v2.1 score of 4 was the best cutoff value for predicting csPCa in patients with lesions only in the transitional zone with respect to the Youden index (0.5896) and negative predictive value (93.10%) with acceptable sensitivity (81.82%) and specificity (77.14%). An adjustment of the cutoff value to 3 for lesions in the peripheral zone would increase the negative predictive value (92.00%) and decrease the false negative rate (2.90%) with an acceptable sensitivity (97.10%) and specificity (30.67%). CONCLUSION: PI-RADS v2.1 score is an effective predictor of csPCa in patients with PSA levels of 4 to 10 ng/mL. Patients with transitional zone or peripheral zone lesions should undergo biopsy if the PI-RADS v2.1 score is ≥ 4 or ≥ 3, respectively.

Predictive model using prostate MRI findings can predict candidates for nerve sparing radical prostatectomy among low-intermediate risk prostate cancer patients.

BACKGROUND: In order to improve postoperative functional outcome, including urinary continence and erectile function, nerve sparing surgery is recommended for patients with clinically localized prostate cancer (PCa). However, due to poor diagnosis accuracy at the preoperative stage, upstaging occurs in a considerable proportion of patients. Multiparametric magnetic resonance imaging (mpMRI) and the Prostate Imaging Reporting and Data System version 2 (PI-RADS v2) have recently shown excellent performance in diagnosis and staging of PCa. The aim of this study was to develop a predictive model based on PI-RADS v2 for postoperative upstaging in patients with low-intermediate risk PCa. METHODS: The medical records of 314 patients with low-intermediate risk PCa [prostate-specific antigen (PSA) level ≤20 ng/mL, Gleason score (GS) <8, and clinical stage < T3] who underwent preoperative mpMRI and radical prostatectomy in the Department of Urology, Peking University First Hospital between January 2012 and July 2019 were reviewed retrospectively. Clinicopathological characteristics were collected. All MRI reports were done at our institution as part of routine clinical practice before prostate biopsy and there was no re-reporting occurred. Using PI-RADS v2, the mpMRI results were assigned to three groups: "negative", "suspicious", and "positive". Multivariate logistic regression analysis was used to assess factors associated with postoperative pathological upstaging, defined as the presence of pT3 at final pathology. A regression coefficient based model for predicting postoperative upstaging was constructed and internally validated using 1,000 bootstrap resamples. The performance of the model was assessed using the area under the receiver operating characteristic curve (AUC). With the optimal cutoff point the performance of the model was assessed through analysis of sensitivity, specificity, positive predictive value, and negative predictive value. RESULTS: Upstaging was observed in 119 (37.9%) patients. The univariate and multivariate analyses revealed that PSA density, biopsy Gleason grade group (GGG), and mpMRI findings were significantly independent predictors for postoperative upstaging (all P<0.05). A predictive model showing very favorable calibration characteristics and higher accuracy than the single variables was constructed (AUC =0.74; P<0.001). At the optimal cutoff point, the model demonstrated a sensitivity and negative predictive value of 87.4% and 87.0%, respectively. CONCLUSIONS: PI-RADS v2 assessment proved to be one of the most valuable predictors for postoperative upstaging in patients with low-intermediate risk PCa. The predictive model, based on PI-RADS v2 assessment, PSA density, and biopsy GGG, may help to select suitable candidates for nerve sparing radical prostatectomy among patients with low-intermediate risk PCa.