Evaluation of extracapsular extension in prostate cancer using qualitative and quantitative multiparametric MRI

Can we predict pathology without surgery? Weighing the added value of multiparametric MRI and whole prostate radiomics in integrative machine learning models

OBJECTIVE

To test the ability of high-performance machine learning (ML) models employing clinical, radiological, and radiomic variables to improve non-invasive prediction of the pathological status of prostate cancer (PCa) in a large, single-institution cohort.

METHODS

Patients who underwent multiparametric MRI and prostatectomy in our institution in 2015-2018 were considered; a total of 949 patients were included. Gradient-boosted decision tree models were separately trained using clinical features alone and in combination with radiological reporting and/or prostate radiomic features to predict pathological T, pathological N, ISUP score, and their change from preclinical assessment. Model behavior was analyzed in terms of performance, feature importance, Shapley additive explanation (SHAP) values, and mean absolute error (MAE). The best model was compared against a naïve model mimicking clinical workflow.

RESULTS

The model including all variables was the best performing (AUC values ranging from 0.73 to 0.96 for the six endpoints). Radiomic features brought a small yet measurable boost in performance, with the SHAP values indicating that their contribution can be critical to successful prediction of endpoints for individual patients. MAEs were lower for low-risk patients, suggesting that the models find them easier to classify. The best model outperformed (p ≤ 0.0001) clinical baseline, resulting in significantly fewer false negative predictions and overall was less prone to under-staging.

CONCLUSIONS

Our results highlight the potential benefit of integrative ML models for pathological status prediction in PCa. Additional studies regarding clinical integration of such models can provide valuable information for personalizing therapy offering a tool to improve non-invasive prediction of pathological status.

CLINICAL RELEVANCE STATEMENT

The best machine learning model was less prone to under-staging of the disease. The improved accuracy of our pathological prediction models could constitute an asset to the clinical workflow by providing clinicians with accurate pathological predictions prior to treatment.

KEY POINTS

• Currently, the most common strategies for pre-surgical stratification of prostate cancer (PCa) patients have shown to have suboptimal performances. • The addition of radiological features to the clinical features gave a considerable boost in model performance. Our best model outperforms the naïve model, avoiding under-staging and resulting in a critical advantage in the clinic. •Machine learning models incorporating clinical, radiological, and radiomics features significantly improved accuracy of pathological prediction in prostate cancer, possibly constituting an asset to the clinical workflow.

Quantitative MRI in the Local Staging of Prostate Cancer: A Systematic Review and Meta-Analysis

BACKGROUND

Local staging of prostate cancer (PCa) is important for treatment planning. Radiologist interpretation using qualitative criteria is variable with high specificity but low sensitivity. Quantitative methods may be useful in the diagnosis of extracapsular extension (ECE).

PURPOSE

To assess the performance of quantitative MRI markers for detecting ECE.

STUDY TYPE

Systematic review and meta-analysis.

SUBJECTS

4800 patients from 28 studies with histopathologically confirmed PCa on radical prostatectomy were pooled for meta-analysis. Patients from 46 studies were included for systematic review.

FIELD STRENGTH/SEQUENCE

Diffusion-weighted, T2-weighted, and dynamic contrast-enhanced MRI at 1.5 T or 3 T.

ASSESSMENT

PubMed, Embase, Web of Science, Scopus, and Cochrane databases were searched to identify studies on diagnostic test accuracy or association of any quantitative MRI markers with ECE. Results extracted by two independent reviewers for tumor contact length (TCL) and mean apparent diffusion coefficient (ADC-mean) were pooled for meta-analysis, but not for other quantitative markers including radiomics due to low number of studies available.

STATISTICAL TESTS

Hierarchical summary receiver operating characteristic (HSROC) curves were computed for both TCL and ADC-mean, but summary operating points were computed for TCL only. Heterogeneity was investigated by meta-regression. Results were significant if P ≤ 0.05.

RESULTS

At the 10 mm threshold for TCL, summary sensitivity and specificity were 0.76 [95% confidence interval (CI) 0.71-0.81] and 0.68 [95% CI 0.63-0.73], respectively. At the 15 mm threshold, summary sensitivity and specificity were 0.70 [95% CI 0.53-0.83] and 0.74 [95% CI 0.60-0.84] respectively. The area under the HSROC curves for TCL and ADC-mean were 0.79 and 0.78, respectively. Significant sources of heterogeneity for TCL included timing of MRI relative to biopsy.

DATA CONCLUSION

Both 10 mm and 15 mm thresholds for TCL may be reasonable for clinical use. From comparison of the HSROC curves, ADC-mean may be superior to TCL at higher sensitivities.

LEVEL OF EVIDENCE

3 TECHNICAL EFFICACY STAGE: 2.

Clinical application of single-shot echo-planar diffusion-weighted imaging with compressed SENSE in prostate MRI at 3T: preliminary experience

OBJECTIVES

Image quality (IQ) of diffusion-weighted imaging (DWI) with single-shot echo-planar imaging (ssEPI) suffers from low signal-to-noise ratio (SNR) in high b-value acquisitions. Compressed SENSE (C-SENSE), which combines SENSE with compressed sensing, enables SNR to be improved by reducing noise. The aim of this study was to compare IQ and prostate cancer (PC) detectability between DWI with ssEPI using SENSE (EPIS) and using C-SENSE (EPICS).

MATERIALS AND METHODS

Twenty-five patients with pathologically proven PC underwent multi-parametric magnetic resonance imaging at 3T. DW images acquired with EPIS and EPICS were assessed for the following: lesion conspicuity (LC), SNR, contrast-to-noise ratio (CNR), mean and standard deviation (SD) of apparent diffusion coefficient (ADC) of lesion (lADCm and lADCsd), coefficient of variation of lesion ADC (lADCcv), and mean ADC of benign prostate (bADCm).

RESULTS

LC were comparable between EPIS and EPICS (p > 0.050), and SNR and CNR were significantly higher in EPICS than EPIS (p = 0.001 and p < 0.001). In both EPIS and EPICS, lADCm was significantly lower than bADCm (p < 0.001). In addition, lADCcv was significantly lower in EPICS than in EPIS (p < 0.001).

CONCLUSION

Compared with EPIS, EPICS has improved IQ and comparable diagnostic performance in PC.

The Quantitative Assessment of Using Multiparametric MRI for Prediction of Extraprostatic Extension in Patients Undergoing Radical Prostatectomy: A Systematic Review and Meta-Analysis

PURPOSE

To investigate the diagnostic performance of using quantitative assessment with multiparametric MRI (mpMRI) for prediction of extraprostatic extension (EPE) in patients with prostate cancer (PCa).

METHODS

We performed a computerized search of MEDLINE, Embase, Cochrane Library, Web of Science, and Google Scholar from inception until July 31, 2021. Summary estimates of sensitivity and specificity were pooled with the bivariate model, and quality assessment of included studies was performed with the Quality Assessment of Diagnostic Accuracy Studies-2. We plotted forest plots to graphically present the results. Multiple subgroup analyses and meta-regression were performed to explore the variate clinical settings and heterogeneity.

RESULTS

A total of 23 studies with 3,931 participants were included. The pooled sensitivity and specificity for length of capsular contact (LCC) were 0.79 (95% CI 0.75-0.83) and 0.77 (95% CI 0.73-0.80), for apparent diffusion coefficient (ADC) were 0.71 (95% CI 0.50-0.86) and 0.71 (95% CI 059-0.81), for tumor size were 0.62 (95% CI 0.57-0.67) and 0.75 (95% CI 0.67-0.82), and for tumor volume were 0.77 (95% CI 0.68-0.84) and 0.72 (95% CI 0.56-0.83), respectively. Substantial heterogeneity was presented among included studies, and meta-regression showed that publication year (≤2017 vs. >2017) was the significant factor in studies using LCC as the quantitative assessment (P=0.02).

CONCLUSION

Four quantitative assessments of LCC, ADC, tumor size, and tumor volume showed moderate to high diagnostic performance of predicting EPE. However, the optimal cutoff threshold varied widely among studies and needs further investigation to establish.

Diagnostic performance of ESUR scoring system for extraprostatic prostate cancer extension: A meta-analysis

PURPOSE

We aimed to evaluate the diagnostic performance of the European Society of Urogenital Radiology (ESUR) scoring system for detection of extraprostatic extension (EPE) in prostate cancer (PCa) by performing a meta-analysis.

MATERIALS AND METHODS

A literature search of MEDLINE, EMBASE, Cochrane Library, Web of Science, and Google Scholar was performed to identify relevant studies from January 2012 to December 2020. We included diagnostic accuracy studies using ESUR scoring system for detection of EPE, and with prostatectomy histopathological results as the reference standard. Quality assessment was performed with the Quality Assessment of Diagnostic Accuracy Studies-2 tool. The summary estimates of sensitivity and specificity were pooled using bivariate random-effects modeling. We conducted multiple subgroup analyses and meta-regression to explore varied clinical settings.

RESULTS

10 studies with a total of 1698 participants were included in this meta-analysis. Pooled sensitivity and specificity were 0.71 (95% CI 0.61-0.80) and 0.76 (95% CI 0.67-0.84), respectively, with the area under ROC of 0.80 (95% CI 0.77-0.84). The Higgins I² statistics demonstrated substantial heterogeneity in both sensitivity (I² = 86.5%) and specificity (I² = 91.6%), meta-regression revealed that the cutoff values (ESUR score ≥ 3 vs. ESUR score ≥ 4, P = 0.02) and malignancy rate (<40% vs. ≥40%, P = 0.04) were significant factors responsible for heterogeneity. Using endorectal coil and higher field strength (3.0 T) showed no additional benefit for EPE detection.

CONCLUSION

The evidence available for ESUR scoring system tends to show moderate diagnostic performance for detection of EPE, and the cutoff values (P = 0.02) and malignancy rate (P = 0.04) were significant factors contributed to the heterogeneity.

Artificial intelligence is a promising prospect for the detection of prostate cancer extracapsular extension with mpMRI: a two-center comparative study

PURPOSE

A balance between preserving urinary continence as well as sexual potency and achieving negative surgical margins is of clinical relevance while implementary difficulty. Accurate detection of extracapsular extension (ECE) of prostate cancer (PCa) is thus crucial for determining appropriate treatment options. We aimed to develop and validate an artificial intelligence (AI)-based tool for detecting ECE of PCa using multiparametric magnetic resonance imaging (mpMRI).

METHODS

Eight hundred and forty nine consecutive PCa patients who underwent mpMRI and prostatectomy without previous radio- or hormonal therapy from two medical centers were retrospectively included. The AI tool was built on a ResNeXt network embedded with a spatial attention map of experts' prior knowledge (PAGNet) from 596 training patients. Model validation was performed in 150 internal and 103 external patients. Performance comparison was made between AI, two experts using a criteria-based ECE grading system, and expert-AI interaction.

RESULTS

An index PAGNet model using a single-slice image yielded the highest areas under the receiver operating characteristic curve (AUC) of 0.857 (95% confidence interval [CI], 0.827-0.884), 0.807 (95% CI, 0.735-0.867), and 0.728 (95% CI, 0.631-0.811) in training, internal, and external validation data, respectively. The performance of two experts (AUC, 0.632 to 0.741 vs 0.715 to 0.857) was lower (paired comparison, all p values < 0.05) than that of AI assessment. When experts' interpretations were adjusted by AI assessments, the performance of two experts was improved.

CONCLUSION

Our AI tool, showing improved accuracy, offers a promising alternative to human experts for ECE staging using mpMRI.

Detecting Muscle Invasion of Bladder Cancer Using a Proposed Magnetic Resonance Imaging Strategy

BACKGROUND

Accurate evaluation of the invasion depth of tumors with a Vesical Imaging-Reporting and Data System (VI-RADS) score of 3 is difficult.

PURPOSE

To evaluate the diagnostic performance of a new magnetic resonance imaging (MRI) strategy based on the integration of the VI-RADS and tumor contact length (TCL) for the diagnosis of muscle-invasive bladder cancer (MIBC).

STUDY TYPE

Single center, retrospective.

SUBJECTS

A group of 179 patients with a mean age of 67 years (range, 24.0-96.0) underwent multiparametric MRI (mpMRI) before surgery, including 147 (82.1%) males and 32 (17.9%) females. Twenty-four (13.4%), 90 (50.3%), 43 (24.0%), 15 (8.4%), and 7 (3.9%) cases were Ta, T1, T2, T3, and T4, respectively.

FIELD STRENGTH/SEQUENCE

A 1.5 T and 3.0 T, T2-weighted turbo spin-echo (TSE), single-shot echo-planar (SS-EPI), diffusion-weighted imaging (DWI), and T1-weighted volumetric interpolated breath-hold examination (T1-VIBE).

ASSESSMENT

Three radiologists independently graded the VI-RADS score and measured the TCL on index lesion images. A proposed MRI strategy called VI-RADS_TCL was introduced by modifying the VI-RADS score, which was downgraded to VI-RADS 3F (equal to a VI-RADS score of 2) if VI-RADS = 3 and TCL < 3 cm.

STATISTICAL TESTS

Intraclass correlation coefficients (ICCs), Mann-Whitney U test, chi-square tests, receiver operating characteristic (ROC) curves, and 2 × 2 contingency tables were applied.

RESULTS

Inter-reader agreement values were 0.941 (95% CI, 0.924-0.955) and 0.934 (95% CI, 0.916-0.948) for the TCL and VI-RADS score. The TCL was significantly increased in the MIBC group (6.40-6.85 cm) compared with the NMIBC group (1.98-2.45 cm) (P < 0.05). The specificity and positive predictive values (PPV) of VI-RADS_TCL were 82.46%-87.72% and 90.91%-91.59%, which were significantly greater than VI-RADS score (P < 0.05). Additionally, 52.17%-55.88% NMIBC lesions with VI-RADS 3 were downgraded to 3F by using VI-RADS_TCL.

DATA CONCLUSION

The proposed MRI strategy could reduce the false-positive rate of lesions with a VI-RADS score of 3 while retaining sensitivity.

EVIDENCE LEVEL

4 TECHNICAL EFFICACY: 2.

Quantitation of bladder cancer for the prediction of muscle layer invasion as a complement to the vesical imaging-reporting and data system

OBJECTIVES

To examine the diagnostic performance of Vesical Imaging-Reporting and Data System (VIRADS) and to find a quantitative indicator for predicting muscle layer invasion of bladder cancer.

METHODS

3-T MRI of 82 patients performed before transurethral resection of bladder tumors or radical cystectomy between July 2018 and June 2019 were retrospectively analyzed. For one index lesion of each patient, two radiologists independently assigned VIRADS score and measured tumor-wall interface (contact length between tumor and bladder wall) on T2-weighted, diffusion-weighted, and dynamic contrast-enhanced MRI. Inter-reader agreement was assessed, and logistic regression analysis was performed to find indicators of muscle layer invasion. Comparison of indicators' diagnostic performance was done with receiver operating characteristic (ROC) curve and generalized linear model analyses. Optimal cutoff point was determined by the Youden index J.

RESULTS

Inter-reader agreement was at least substantial for VIRADS categorization (κ 0.77-0.81), and almost perfect for tumor-wall interface (intraclass correlation coefficient 0.88-0.90). Tumor-wall interface (odds ratio [OR] 1.90-2.00) and VIRADS score (OR 8.59-8.89) were independently associated with muscle layer invasion (p ≤ 0.02). For VIRADS, area under the ROC curve (AUROC) was 0.94, and the accuracy was 0.93 at score 3, the optimal threshold for predicting muscle layer invasion. Depending on the MRI sequence, tumor-wall interface showed AUROCs of 0.90-0.92 and accuracy of 0.84-0.90 at suggested thresholds (3 ± 0.3 cm). Tumor-wall interface showed insignificant differences in accuracy compared with VIRADS (p > 0.10), except as measured on diffusion-weighted images (p = 0.01).

CONCLUSIONS

VIRADS is a good predictor of muscle layer invasion. As an independent quantitative indicator, tumor-wall interface may complement VIRADS to enhance prediction.

KEY POINTS

• Vesical Imaging-Reporting and Data System (VIRADS) is a promising predictor of muscle invasion of bladder cancer with good reproducibility, as suggested by previous studies. • VIRADS score and the tumor-wall interface (curvilinear contact length between the tumor and the bladder wall) are independent predictors of muscle layer invasion. • As an easy-to-use quantitative indicator, tumor-wall interface is expected to be used as an indicator complementary to VIRADS, a qualitative indicator.

MRI detected extaprostatic extension (EPE) in prostate cancer: Do all T3a patients have the same outcomes?

MRI-detected T3a prostate cancer is a heterogeneous disease. This post-hoc analysis of a prospective trial found that patients with T3a disease presenting obliteration of the recto-prostatic angle, contact-asymmetry of neuro-vascular bundle and periprostatic fat invasion, may be at higher risk of biochemical failure and metastases.

Radiomics Based on Multiparametric Magnetic Resonance Imaging to Predict Extraprostatic Extension of Prostate Cancer

Background: To develop a radiomics model based on multiparametric MRI (mpMRI) for preoperative prediction of extraprostatic extension (EPE) in patients with prostate cancer (PCa). Methods: Ninety-five pathology-confirmed PCa patients with 115 lesions (49 positive and 66 negative) were retrospectively enrolled. A 3.0T MR scanner was used to perform T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and dynamic contrast-enhanced imaging (DCE). Radiomics features extracted from T2WI, DWI, apparent diffusion coefficient (ADC), and DCE were used to build a radiomics model. Patients' clinical and pathological variables were also obtained to build a clinical model. The radiomics model and clinical model were further integrated to build a combined nomogram. All lesions were randomly divided into the training group (82 lesions) and the validation group (33 lesions). A least absolute shrinkage and selection operator (LASSO) regression algorithm was applied to build the radiomics model. The diagnostic performance of different models was assessed by calculating the area under the curve (AUC) and compared using the Delong test. The calibration curve and decision curve analyses were used to assess the calibration and clinical usefulness of the radiomics model. Results: The AUC values for the radiomics model in the training and validation group were 0.919 and 0.865, respectively, with a good calibration performance. The decision curve analysis confirmed the clinical utility of the radiomics model. The accuracy, sensitivity, and specificity were 81.8, 71.4, and 89.5% in the validation group. In the validation group, the radiomics model outperformed the clinical model (AUC = 0.658, P = 0.020), and was comparable with the combined nomogram (AUC = 0.857, P = 0.644). Conclusion: The radiomics model based on mpMRI could different EPE and non-EPE lesions with satisfactory diagnostic performance, and this model might assist in predicting EPE before prostatectomy.

Prognostic Significance for Long-Term Outcomes Following Radical Prostatectomy in Men with Prostate Cancer: Evaluation with Prostate Imaging Reporting and Data System Version 2

Objective

To retrospectively determine whether the use of the Prostate Imaging Reporting and Data System (PI-RADS) version 2 (v2) helps predict long-term outcomes for prostate cancer (PCa) patients following radical prostatectomy (RP).

Materials and Methods

A total of 166 patients with localized PCa evaluated with multiparametric magnetic resonance imaging (mpMRI) at 3T before RP were enrolled. Three groups were created based on PI-RADS v2 score used to predict clinical outcomes: group A, ≥ 3; group B, ≥ 4; group C, 5. We calculated biochemical recurrence-free survival (RFS) and progression-free survival (PFS). Cox proportion hazards models were used to identify variables predictive of biochemical recurrence and disease progression.

Results

During a median follow-up of 9.1 years, biochemical recurrence occurred in 67 patients (40.4%) and disease progression occurred in 55 patients (33.1%). In all groups, 10-year RFS and 10-year PFS were significantly lower for PI-RADS scores ≥ 3, ≥ 4 and 5 than for score < 3, < 4 and < 5 (p <0.05), respectively. In multivariate analysis, PI-RADS score ≥ 3 and score 5 were significant independent risk marker for biochemical recurrence (hazard ratio [HR] = 5.58, p = 0.018; HR = 1.75, p = 0.033) and disease progression (HR = 3.99, p = 0.047; HR = 2.31, p = 0.040). Moderate inter-observer agreement was seen for PI-RADS scoring.

Conclusion

PI-RADS v2 may be used to predict long-term outcomes following RP in PCa.

Accuracy of multiparametric magnetic resonance imaging for detecting extracapsular extension in prostate cancer: a systematic review and meta-analysis

OBJECTIVE

To evaluate the diagnostic accuracy of multiparametric MRI (mpMRI) for detecting extracapsular extension (ECE) in patients with prostate cancer (PCa).

METHODS AND MATERIALS

We searched MEDLINE, PubMed, Embase and the Cochrane library up to December 2018. We included studies that used mpMRI to differentiate ECE from organ-confined PCa with a combination of T₂ weighted imaging (T₂WI), diffusion-weighted imaging, and dynamic contrast-enhanced MRI. All studies included had pathological diagnosis with radical prostatectomy. Two reviewers independently assessed the methodological quality of included studies by using Quality Assessment of Diagnostic Accuracy Studies 2 tool. We calculated pooled sensitivity, specificity, positive and negative predictive values, diagnostic odds ratios and receiver operating characteristic curve for mpMRI from 2 × 2 tables.

RESULTS

A total of 17 studies that comprised 3374 participants were included. The pooled data showed a sensitivity of 0.55 (95% confidence interval 0.43, 0.66]) and specificity of 0.87 (95% confidence interval 0.82, 0.91) for extracapsular extension detection in PCa.

CONCLUSION

First, our meta-analysis shows moderate sensitivity and high specificity for mpMRI to differentiate ECE from organ-confined prostate cancer before surgery. Second, our meta-analysis shows that mpMRI had no significant differences in performance compared with the former meta-analysis with use of T₂WI alone or with additional functional MRI.

ADVANCES IN KNOWLEDGE

It is the first meta-analysis to evaluate the accuracy of mpMRI in combination of TWI, diffusion-weightedimaging and dynamiccontrast-enhanced-MRI for extracapsular extension detection.

Correlations between Apparent Diffusion Coefficient and Gleason Score in Prostate Cancer: A Systematic Review

BACKGROUND

Reported data regarding the associations between apparent diffusion coefficient (ADC) of diffusion-weighted imaging (DWI) and Gleason score in prostate cancer (PC) are inconsistent.

OBJECTIVE

The aim of the present systematic review was to analyze relationships between ADC and Gleason score in PC.

DESIGN, SETTING, AND PARTICIPANTS

MEDLINE library, SCOPUS, and EMBASE databases were screened for relationships between ADC and Gleason score in PC up to April 2018. Overall, 39 studies with 2457 patients were identified. Data on the following parameters were extracted from the literature: number of patients, cancer localization, and correlation coefficients between ADC and Gleason score.

OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS

Associations between ADC and Gleason score were analyzed by the Spearman's correlation coefficient.

RESULTS AND LIMITATIONS

In overall sample, the pooled correlation coefficient between ADC and Gleason score was -0.45 (95% confidence interval [CI]=[-0.50; -0.40]). In PC in the transitional zone, the pooled correlation coefficient was -0.22 (95% CI=[-0.47; 0.03]). In PC in the peripheral zone, the pooled correlation coefficient was -0.48 (95% CI=[-0.54; -0.42]).

CONCLUSIONS

In PC located in the peripheral zone, ADC correlated moderately with Gleason score. In PC located in the transitional zone, ADC correlated weakly with Gleason score.

PATIENT SUMMARY

We reviewed studies using apparent diffusion coefficient for the prediction of Gleason score in prostate cancer patients.

[PI-RADS 2.0 for Prostate MRI]

The PI-RADS 2.0 classification was established by an international collaboration of the European Society of Urogenital Radiology (ESUR), the American College of Radiology (ACR), and AdMetech Foundation to globally standardize the acquisition and interpretation of multiparametric prostate magnetic resonance imaging (MRI). The PI-RADS 2.0 aims to improve the detection, localization, staging and risk stratification of patients with suspected or histologically confirmed prostate cancer. Suspicious areas on T2-weighted (T2w) MRI, diffusion-weighted MRI (DWI) and dynamic contrast-enhanced MRI (DCE-MRI) were assessed for the presence of a clinically significant prostate cancer with scale from 1 to 5 with 5 being most likely to represent clinically significant prostate cancer. The dominant sequence to detect significant prostate cancer in the peripheral zone is DWI and for the transition zone T2w images. For the local staging of prostate cancer criteria for the assessment of an extracapsular growth were formulated.

[MRI of the prostate]

New clinical and technological advances in the field of magnetic resonance imaging (MRI) and targeted image-guided biopsy techniques have significantly improved the detection, localization and staging as well as active surveillance of prostate cancer in recent years. Multiparametric MRI (mpMRI) is currently the main imaging technique for the detection, characterization and diagnostics of metastasizing prostate cancer and is of high diagnostic importance for local staging within the framework of the detection of prostate cancer.