PI-QUAL version 2: an update of a standardised scoring system for the assessment of image quality of prostate MRI.

Multiparametric MRI is the optimal primary investigation when prostate cancer is suspected, and its ability to rule in and rule out clinically significant disease relies on high-quality anatomical and functional images. Avenues for achieving consistent high-quality acquisitions include meticulous patient preparation, scanner setup, optimised pulse sequences, personnel training, and artificial intelligence systems. The impact of these interventions on the final images needs to be quantified. The prostate imaging quality (PI-QUAL) scoring system was the first standardised quantification method that demonstrated the potential for clinical benefit by relating image quality to cancer detection ability by MRI. We present the updated version of PI-QUAL (PI-QUAL v2) which applies to prostate MRI performed with or without intravenous contrast medium using a simplified 3-point scale focused on critical technical and qualitative image parameters. CLINICAL RELEVANCE STATEMENT: High image quality is crucial for prostate MRI, and the updated version of the PI-QUAL score (PI-QUAL v2) aims to address the limitations of version 1. It is now applicable to both multiparametric MRI and MRI without intravenous contrast medium. KEY POINTS: High-quality images are essential for prostate cancer diagnosis and management using MRI. PI-QUAL v2 simplifies image assessment and expands its applicability to prostate MRI without contrast medium. PI-QUAL v2 focuses on critical technical and qualitative image parameters and emphasises T2-WI and DWI.

PI-RADS Upgrading Rules: Impact on Prostate Cancer Detection and Biopsy Avoidance of MRI-Directed Diagnostic Pathways.

BACKGROUND. PI-RADS incorporates rules by which ancillary sequence findings upgrade a dominant score to a higher final category. Evidence on the upgrading rules' impact on diagnostic pathways remains scarce. OBJECTIVE. The purpose of this article was to evaluate the clinical net benefit of the PI-RADS upgrading rules in MRI-directed diagnostic pathways. METHODS. This study was a retrospective analysis of a prospectively maintained clinical registry. The study included patients without known prostate cancer who underwent prostate MRI followed by prostate biopsy from January 2016 to May 2020. Clinically significant prostate cancer (csPCa) was defined as International Society of Urological Pathology (ISUP) grade group 2 and higher. csPCa detection was compared between dominant (i.e., no upgrade rule applied) and upgraded lesions. Decision-curve analysis was used to compare the net benefit, considering the trade-off of csPCa detection and biopsy avoidance, of MRI-directed pathways in scenarios considering and disregarding PI-RADS upgrading rules. These included a biopsy-all pathway, MRI-focused pathway (no biopsy for PI-RADS ≤ 2), and risk-based pathway (use of PSA density ≥ 0.15 ng/mL2 to select patients with PI-RADS ≤ 3 for biopsy). RESULTS. The sample comprised 716 patients (mean age, 64.9 years; 93 with a PI-RADS ≤ 2 examination, 623 with total of 780 PI-RADS ≥ 3 lesions). Frequencies of csPCa were not significantly different between dominant and upgraded PI-RADS 3 transition zone lesions (20% vs 19%, respectively), dominant and upgraded PI-RADS 4 transition zone lesions (33% vs 26%), and dominant and upgraded PI-RADS 4 peripheral zone lesions (58% vs 45%) (p > .05). In the biopsy-all, per-guideline MRI-focused, MRI-focused disregarding upgrading rules, per-guideline risk-based, and risk-based disregarding upgrading rules pathways, csPCa frequency was 53%, 52%, 51%, 52%, and 48% and biopsy avoidance was 0%, 13%, 16%, 19%, and 25%, respectively. Disregarding upgrading rules yielded 5.5 and 1.9 biopsies avoided per missed csPCa for MRI-focused and risk-based pathways, respectively. At probability thresholds for biopsy selection of 7.5-30.0%, net benefit was highest for the per-guideline risk-based pathway. CONCLUSION. Disregarding PI-RADS upgrading rules reduced net clinical bene fit of the risk-based MRI-directed diagnostic pathway when considering trade-offs between csPCa detection and biopsy avoidance. CLINICAL IMPACT. This study supports the application of PI-RADS upgrading rules to optimize biopsy selection, particularly in risk-based pathways.

New Prostate MRI Scoring Systems (PI-QUAL, PRECISE, PI-RR, and PI-FAB): AJR Expert Panel Narrative Review.

Multiparametric MRI (mpMRI), interpreted using PI-RADS, improves the initial detection of clinically significant prostate cancer (PCa). Prostate MR image quality has increasingly recognized relevance to the use of mpMRI for PCa diagnosis. Additionally, mpMRI is increasingly used in scenarios beyond initial detection, including active surveillance and assessment for local recurrence after prostatectomy, radiation therapy, or focal therapy. Acknowledging these evolving demands, specialized prostate MRI scoring systems beyond PI-RADS have emerged, to address distinct scenarios and unmet needs. Examples include Prostate Imaging Quality (PI-QUAL) for assessment of image quality of mpMRI, Prostate Cancer Radiologic Estimation of Change in Sequential Evaluation (PRECISE) recommendations for evaluation of serial mpMRI examinations during active surveillance, Prostate Imaging for Recurrence Reporting System (PI-RR) for assessment for local recurrence after prostatectomy or radiation therapy, and Prostate Imaging after Focal Ablation (PI-FAB) for assessment for local recurrence after focal therapy. These systems' development and early uptake signal a compelling shift towards prostate MRI standardization in different scenarios, and ongoing research will help refine their roles in practice. This AJR Expert Panel Narrative Review critically examines these new prostate MRI scoring systems (PI-QUAL, PRECISE, PI-RR, and PI-FAB), analyzing the available evidence, delineating current limitations, and proposing solutions for improvement.

PI-RADS: Where Next?

Prostate MRI plays an important role in the clinical management of localized prostate cancer, mainly assisting in biopsy decisions and guiding biopsy procedures. The Prostate Imaging Reporting and Data System (PI-RADS) has been available to radiologists since 2012, with the most up-to-date and actively used version being PI-RADS version 2.1. This review article discusses the current use of PI-RADS, including its limitations and controversies, and summarizes research that aims to improve future iterations of this system.

Impact of Prostate Urethral Lift Device on Prostate Magnetic Resonance Image Quality.

PURPOSE: Prostatic urethral lift with UroLift is a minimally invasive approach to treat symptomatic benign prostatic hypertrophy. This device causes artifacts on prostate magnetic resonance images. Our aim was to evaluate the impact of artifact on prostate magnetic resonance image quality. MATERIALS AND METHODS: This was a single-center retrospective review of patients with UroLift who subsequently had prostate magnetic resonance imaging. Two readers graded UroLift artifact on each pulse sequence using a 5-point scale (1-nondiagnostic; 5-no artifact). Prostate Imaging Quality scores were assigned for the whole data set. The volume of gland obscured by artifact was measured. Linear and logistic regression models were used to identify predictors of poor image quality. RESULTS: Thirty-seven patients were included. Poor image quality occurs more in the transition zone than the peripheral zone (15% vs 3%), at base/mid regions vs the apex (13%, 9%, and 5%, respectively) and on diffusion-weighted images vs T2-weighted and dynamic contrast-enhanced sequences (27%, 0.3%, 0%, respectively; P < .001). Suboptimal image quality (ie, Prostate Imaging Quality score <2) was found in 16%-24% of exams. The percentage of gland obscured by the UroLift artifact was higher on diffusion-weighted images and dynamic contrast-enhanced sequences than T2-weighted (32%, 9%, and 6%, respectively; P < .001). CONCLUSIONS: UroLift artifact negatively affects prostate magnetic resonance image quality with greater impact in the mid-basal transition zone, obscuring a third of the gland on diffusion-weighted images. Patients considering this procedure should be counseled on the impact of this device on image quality and its potential implications for any image-guided prostate cancer workup.

MRI assessment of rectal cancer response to neoadjuvant therapy: a multireader study.

OBJECTIVES: A watch and wait strategy with the goal of organ preservation is an emerging treatment paradigm for rectal cancer following neoadjuvant treatment. However, the selection of appropriate patients remains a challenge. Most previous efforts to measure the accuracy of MRI in assessing rectal cancer response used a small number of radiologists and did not report variability among them. METHODS: Twelve radiologists from 8 institutions assessed baseline and restaging MRI scans of 39 patients. The participating radiologists were asked to assess MRI features and to categorize the overall response as complete or incomplete. The reference standard was pathological complete response or a sustained clinical response for > 2 years. RESULTS: We measured the accuracy and described the interobserver variability of interpretation of rectal cancer response between radiologists at different medical centers. Overall accuracy was 64%, with a sensitivity of 65% for detecting complete response and specificity of 63% for detecting residual tumor. Interpretation of the overall response was more accurate than the interpretation of any individual feature. Variability of interpretation was dependent on the patient and imaging feature investigated. In general, variability and accuracy were inversely correlated. CONCLUSIONS: MRI-based evaluation of response at restaging is insufficiently accurate and has substantial variability of interpretation. Although some patients' response to neoadjuvant treatment on MRI may be easily recognizable, as seen by high accuracy and low variability, that is not the case for most patients. KEY POINTS: • The overall accuracy of MRI-based response assessment is low and radiologists differed in their interpretation of key imaging features. • Some patients' scans were interpreted with high accuracy and low variability, suggesting that these patients' pattern of response is easier to interpret. • The most accurate assessments were those of the overall response, which took into consideration both T2W and DWI sequences and the assessment of both the primary tumor and the lymph nodes.

Diagnostic Accuracy and Observer Agreement of the MRI Prostate Imaging for Recurrence Reporting Assessment Score.

Background Prostate cancer local recurrence location and extent must be determined in an accurate and timely manner. Because of the lack of a standardized MRI approach after whole-gland treatment, a panel of international experts recently proposed the Prostate Imaging for Recurrence Reporting (PI-RR) assessment score. Purpose To determine the diagnostic accuracy of PI-RR for detecting local recurrence in patients with biochemical recurrence (BCR) after radiation therapy (RT) or radical prostatectomy (RP) and to evaluate the interreader variability of PI-RR scoring. Materials and Methods This retrospective observational study included patients who underwent multiparametric MRI between September 2016 and May 2021 for BCR after RT or RP. MRI scans were analyzed, and a PI-RR score was assigned independently by four radiologists. The reference standard was defined using histopathologic findings, follow-up imaging, or clinical response to treatment. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated to assess PI-RR performance for each reader. The intraclass correlation coefficient was used to determine interreader agreement. Results A total of 100 men were included: 48 patients after RT (median age, 76 years [IQR, 70-82 years]) and 52 patients after RP (median age, 70 years [IQR, 66-74 years]). After RT, with PI-RR of 3 or greater as a cutoff (assigned when recurrence is uncertain), diagnostic performance ranges were 71%-81% sensitivity, 74%-93% specificity, 71%-89% PPV, 79%-86% NPV, and 77%-88% accuracy across the four readers. After RP, with PI-RR of 3 or greater as a cutoff, performance ranges were 59%-83% sensitivity, 87%-100% specificity, 88%-100% PPV, 66%-80% NPV, and 75%-85% accuracy. The intraclass correlation coefficient was 0.87 across the four readers for both the RT and RP groups. Conclusion MRI scoring with the Prostate Imaging for Recurrence Reporting assessment provides structured, reproducible, and accurate evaluation of local recurrence after definitive therapy for prostate cancer. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Haider in this issue. An earlier incorrect version appeared online. This article was corrected on May 11, 2022.

Prostate MRI Qualification: AJR Expert Panel Narrative Review.

Prostate MRI is now established as a first-line investigation for individuals presenting with suspected localized or locally advanced prostate cancer. Successful delivery of the MRI-directed pathway for prostate cancer diagnosis relies on high-quality imaging as well as the interpreting radiologist's experience and expertise. Radiologist certification in prostate MRI may help limit interreader variability, optimize outcomes, and provide individual radiologists with documentation of meeting predefined standards. This AJR Expert Panel Narrative Review summarizes existing certification proposals, recognizing variable progress across regions in establishing prostate MRI certification programs. To our knowledge, Germany is the only country with a prostate MRI certification process that is currently available for radiologists. However, prostate MRI certification programs have also recently been proposed in the United States and United Kingdom and by European professional society consensus panels. Recommended qualification processes entail a multifaceted approach, incorporating components such as minimum case numbers, peer learning, course participation, continuing medical education credits, and feedback from pathology results. Given the diversity in health care systems, including in the provision and availability of MRI services, national organizations will likely need to take independent approaches to certification and accreditation. The relevant professional organizations should begin developing these programs or continue existing plans for implementation.

The utility of laparoscopic ultrasound during minimally invasive liver procedures in patients with malignant liver tumors who have undergone preoperative magnetic resonance imaging.

BACKGROUND: The aim of this study was to assess the utility of laparoscopic ultrasound (LUS) during minimally invasive liver procedures in patients with malignant liver tumors who underwent preoperative magnetic resonance imaging (MRI). METHODS: Medical records of patients with malignant liver lesions who underwent laparoscopic liver surgery between October 2005 and January 2018 and who underwent an MRI examination at our institution within a month before surgery were collected from a prospectively maintained database. The size and location of tumors detected on LUS, as well as whether they were seen on preoperative imaging, were recorded. Univariate and multivariate regression analyses were performed to identify factors that were associated with the detection of liver lesions on LUS that were not seen on preoperative MRI. RESULTS: A total of 467 lesions were identified in 147 patients. Tumor types included colorectal cancer metastasis (n = 53), hepatocellular cancer (n = 38), neuroendocrine metastasis (n = 23), and others (n = 33). Procedures included ablation (67%), resection (23%), combined resection and ablation (6%), and diagnostic laparoscopy with biopsy (4%). LUS identified 39 additional lesions (8.4%) that were not seen on preoperative MRI in 14 patients (10%). These were colorectal cancer (n = 20, 51%), neuroendocrine (n = 11, 28%) and other metastases (n = 8, 21%). These additional findings on LUS changed the treatment plan in 13 patients (8.8%). Factors predicting tumor detection on LUS but not on MRI included obesity (p = 0.02), previous exposure to chemotherapy (p < 0.001), and lesion size < 1 cm (p < 0.001). CONCLUSION: This study demonstrates that, despite advances in MRI, LUS performed during minimally invasive liver procedures may detect additional tumors in 10% of patients with liver malignancies, with the highest yield seen in obese patients with previous exposure to chemotherapy. These results support the routine use of LUS by hepatic surgeons.

Influence of 5-Alpha Reductase Inhibitors on Prostate Cancer Detection with Magnetic Resonance Imaging: A Matched Cohort Study.

PURPOSE: We evaluated the influence of 5-alpha reductase inhibitors (5-ARIs) on the performance of magnetic resonance imaging (MRI) for detection of Gleason grade group (GG) ≥2 prostate cancer, and on apparent diffusion coefficient (ADC) maps. MATERIALS AND METHODS: This single center, retrospective study included men who had MRI for initial detection or active surveillance of prostate cancer. The study group included 59 men who used for 5-ARIs for ≥12 months, and the control group included 59 men who were matched for both MRI indication and biopsy results. DeLong's test was used to compare the area under the receiver operating characteristic curve (AUC) for detection of GG ≥2 cancer between the groups. Wilcoxon rank sum test was used for comparison of lesions apparent diffusion coefficient (ADC) metrics between the groups. RESULTS: MRI accuracy in the study group (AUC=0.778) was not significantly different compared to the control group (AUC=0.821; 95% CI for difference 0.22-0.13; p=0.636). In the control group, all ADC metrics were lower in lesions with GG ≥2 cancer on biopsy than in those with GG 1 cancer or negative results (p=0.001-0.01). In the study group, this difference was significant only when the mean ADC of the lesions was normalized by the ADC of urine (p=0.044). CONCLUSIONS: Long-term exposure to 5-ARIs does not seem to impair the detection of significant cancer on MRI but may affect the ability of ADC metrics to discriminate between lesions that harbor significant cancer and those that harbor insignificant cancer or benign tissue.

PI-RADS Version 2.1: A Critical Review, From the AJR Special Series on Radiology Reporting and Data Systems.

PI-RADS version 2.1 updates the technical parameters for multiparametric MRI (mpMRI) of the prostate and revises the imaging interpretation criteria while maintaining the framework introduced in version 2. These changes have been considered an improvement, although some issues remain unresolved, and new issues have emerged. Areas for improvement discussed in this review include the need for more detailed mpMRI protocols with optimization for 1.5-T and 3-T systems; lack of validation of revised transition zone interpretation criteria and need for clarifications of the revised DWI and dynamic contrast-enhanced imaging criteria and central zone (CZ) assessment; the need for systematic evaluation and reporting of background changes in signal intensity in the prostate that can negatively affect cancer detection; creation of a new category for lesions that do not fit into the PI-RADS assessment categories (i.e., PI-RADS M category); inclusion of quantitative parameters beyond size to evaluate lesion aggressiveness; adjustments to the structured report template, including standardized assessment of the risk of extraprostatic extension; development of parameters for image quality and performance control; and suggestions for expansion of the system to other indications (e.g., active surveillance and recurrence).

Radiomic Features of Primary Rectal Cancers on Baseline T2 -Weighted MRI Are Associated With Pathologic Complete Response to Neoadjuvant Chemoradiation: A Multisite Study.

BACKGROUND: Twenty-five percent of rectal adenocarcinoma patients achieve pathologic complete response (pCR) to neoadjuvant chemoradiation and could avoid proctectomy. However, pretreatment clinical or imaging markers are lacking in predicting response to chemoradiation. Radiomic texture features from MRI have recently been associated with therapeutic response in other cancers. PURPOSE: To construct a radiomics texture model based on pretreatment MRI for identifying patients who will achieve pCR to neoadjuvant chemoradiation in rectal cancer, including validation across multiple scanners and sites. STUDY TYPE: Retrospective. SUBJECTS: In all, 104 rectal cancer patients staged with MRI prior to long-course chemoradiation followed by proctectomy; curated from three institutions. FIELD STRENGTH/SEQUENCE: 1.5T-3.0T, axial higher resolution T2 -weighted turbo spin echo sequence. ASSESSMENT: Pathologic response was graded on postsurgical specimens. In total, 764 radiomic features were extracted from single-slice sections of rectal tumors on processed pretreatment T2 -weighted MRI. STATISTICAL TESTS: Three feature selection schemes were compared for identifying radiomic texture descriptors associated with pCR via a discovery cohort (one site, N = 60, cross-validation). The top-selected radiomic texture features were used to train and validate a random forest classifier model for pretreatment identification of pCR (two external sites, N = 44). Model performance was evaluated via area under the curve (AUC), accuracy, sensitivity, and specificity. RESULTS: Laws kernel responses and gradient organization features were most associated with pCR (P ≤ 0.01); as well as being commonly identified across all feature selection schemes. The radiomics model yielded a discovery AUC of 0.699 ± 0.076 and a hold-out validation AUC of 0.712 with 70.5% accuracy (70.0% sensitivity, 70.6% specificity) in identifying pCR. Radiomic texture features were resilient to variations in magnetic field strength as well as being consistent between two different expert annotations. Univariate analysis revealed no significant associations of baseline clinicopathologic or MRI findings with pCR (P = 0.07-0.96). DATA CONCLUSION: Radiomic texture features from pretreatment MRIs may enable early identification of potential pCR to neoadjuvant chemoradiation, as well as generalize across sites. LEVEL OF EVIDENCE: 3 TECHNICAL EFFICACY STAGE: 2.

Data Augmentation and Transfer Learning to Improve Generalizability of an Automated Prostate Segmentation Model.

OBJECTIVE. Deep learning applications in radiology often suffer from overfitting, limiting generalization to external centers. The objective of this study was to develop a high-quality prostate segmentation model capable of maintaining a high degree of performance across multiple independent datasets using transfer learning and data augmentation. MATERIALS AND METHODS. A retrospective cohort of 648 patients who underwent prostate MRI between February 2015 and November 2018 at a single center was used for training and validation. A deep learning approach combining 2D and 3D architecture was used for training, which incorporated transfer learning. A data augmentation strategy was used that was specific to the deformations, intensity, and alterations in image quality seen on radiology images. Five independent datasets, four of which were from outside centers, were used for testing, which was conducted with and without fine-tuning of the original model. The Dice similarity coefficient was used to evaluate model performance. RESULTS. When prostate segmentation models utilizing transfer learning were applied to the internal validation cohort, the mean Dice similarity coefficient was 93.1 for whole prostate and 89.0 for transition zone segmentations. When the models were applied to multiple test set cohorts, the improvement in performance achieved using data augmentation alone was 2.2% for the whole prostate models and 3.0% for the transition zone segmentation models. However, the best test-set results were obtained with models fine-tuned on test center data with mean Dice similarity coefficients of 91.5 for whole prostate segmentation and 89.7 for transition zone segmentation. CONCLUSION. Transfer learning allowed for the development of a high-performing prostate segmentation model, and data augmentation and fine-tuning approaches improved performance of a prostate segmentation model when applied to datasets from external centers.

Multicenter Multireader Evaluation of an Artificial Intelligence-Based Attention Mapping System for the Detection of Prostate Cancer With Multiparametric MRI.

OBJECTIVE. The purpose of this study was to evaluate in a multicenter dataset the performance of an artificial intelligence (AI) detection system with attention mapping compared with multiparametric MRI (mpMRI) interpretation in the detection of prostate cancer. MATERIALS AND METHODS. MRI examinations from five institutions were included in this study and were evaluated by nine readers. In the first round, readers evaluated mpMRI studies using the Prostate Imaging Reporting and Data System version 2. After 4 weeks, images were again presented to readers along with the AI-based detection system output. Readers accepted or rejected lesions within four AI-generated attention map boxes. Additional lesions outside of boxes were excluded from detection and categorization. The performances of readers using the mpMRI-only and AI-assisted approaches were compared. RESULTS. The study population included 152 case patients and 84 control patients with 274 pathologically proven cancer lesions. The lesion-based AUC was 74.9% for MRI and 77.5% for AI with no significant difference (p = 0.095). The sensitivity for overall detection of cancer lesions was higher for AI than for mpMRI but did not reach statistical significance (57.4% vs 53.6%, p = 0.073). However, for transition zone lesions, sensitivity was higher for AI than for MRI (61.8% vs 50.8%, p = 0.001). Reading time was longer for AI than for MRI (4.66 vs 4.03 minutes, p < 0.001). There was moderate interreader agreement for AI and MRI with no significant difference (58.7% vs 58.5%, p = 0.966). CONCLUSION. Overall sensitivity was only minimally improved by use of the AI system. Significant improvement was achieved, however, in the detection of transition zone lesions with use of the AI system at the cost of a mean of 40 seconds of additional reading time.

RadioGraphics Update: PI-RADS Version 2.1-A Pictorial Update.

Editor's Note.-Articles in the RadioGraphics Update section provide current knowledge to supplement or update information found in full-length articles previously published in RadioGraphics. Authors of the previously published article provide a brief synopsis that emphasizes important new information such as technological advances, revised imaging protocols, new clinical guidelines involving imaging, or updated classification schemes. Articles in this section are published solely online and are linked to the original article.

Correlation between MRI phenotypes and a genomic classifier of prostate cancer: preliminary findings.

OBJECTIVES: We sought to evaluate the correlation between MRI phenotypes of prostate cancer as defined by PI-RADS v2 and the Decipher Genomic Classifier (used to estimate the risk of early metastases). METHODS: This single-center, retrospective study included 72 nonconsecutive men with prostate cancer who underwent MRI before radical prostatectomy performed between April 2014 and August 2017 and whose MRI registered lesions were microdissected from radical prostatectomy specimens and then profiled using Decipher (89 lesions; 23 MRI invisible [PI-RADS v2 scores ≤ 2] and 66 MRI visible [PI-RADS v2 scores ≥ 3]). Linear regression analysis was used to assess clinicopathologic and MRI predictors of Decipher results; correlation coefficients (r) were used to quantify these associations. AUC was used to determine whether PI-RADS v2 could accurately distinguish between low-risk (Decipher score < 0.45) and intermediate-/high-risk (Decipher score ≥ 0.45) lesions. RESULTS: MRI-visible lesions had higher Decipher scores than MRI-invisible lesions (mean difference 0.22; 95% CI 0.13, 0.32; p < 0.0001); most MRI-invisible lesions (82.6%) were low risk. PI-RADS v2 had moderate correlation with Decipher (r = 0.54) and had higher accuracy (AUC 0.863) than prostate cancer grade groups (AUC 0.780) in peripheral zone lesions (95% CI for difference 0.01, 0.15; p = 0.018). CONCLUSIONS: MRI phenotypes of prostate cancer are positively correlated with Decipher risk groups. Although PI-RADS v2 can accurately distinguish between lesions classified by Decipher as low or intermediate/high risk, some lesions classified as intermediate/high risk by Decipher are invisible on MRI. KEY POINTS: • MRI phenotypes of prostate cancer as defined by PI-RADS v2 positively correlated with a genomic classifier that estimates the risk of early metastases. • Most but not all MRI-invisible lesions had a low risk for early metastases according to the genomic classifier. • MRI could be used in conjunction with genomic assays to identify lesions that may carry biological potential for early metastases.

Comparing radiomic classifiers and classifier ensembles for detection of peripheral zone prostate tumors on T2-weighted MRI: a multi-site study.

BACKGROUND: For most computer-aided diagnosis (CAD) problems involving prostate cancer detection via medical imaging data, the choice of classifier has been largely ad hoc, or been motivated by classifier comparison studies that have involved large synthetic datasets. More significantly, it is currently unknown how classifier choices and trends generalize across multiple institutions, due to heterogeneous acquisition and intensity characteristics (especially when considering MR imaging data). In this work, we empirically evaluate and compare a number of different classifiers and classifier ensembles in a multi-site setting, for voxel-wise detection of prostate cancer (PCa) using radiomic texture features derived from high-resolution in vivo T2-weighted (T2w) MRI. METHODS: Twelve different supervised classifier schemes: Quadratic Discriminant Analysis (QDA), Support Vector Machines (SVMs), naïve Bayes, Decision Trees (DTs), and their ensemble variants (bagging, boosting), were compared in terms of classification accuracy as well as execution time. Our study utilized 85 prostate cancer T2w MRI datasets acquired from across 3 different institutions (1 for discovery, 2 for independent validation), from patients who later underwent radical prostatectomy. Surrogate ground truth for disease extent on MRI was established by expert annotation of pre-operative MRI through spatial correlation with corresponding ex vivo whole-mount histology sections. Classifier accuracy in detecting PCa extent on MRI on a per-voxel basis was evaluated via area under the ROC curve. RESULTS: The boosted DT classifier yielded the highest cross-validated AUC (= 0.744) for detecting PCa in the discovery cohort. However, in independent validation, the boosted QDA classifier was identified as the most accurate and robust for voxel-wise detection of PCa extent (AUCs of 0.735, 0.683, 0.768 across the 3 sites). The next most accurate and robust classifier was the single QDA classifier, which also enjoyed the advantage of significantly lower computation times compared to any of the other methods. CONCLUSIONS: Our results therefore suggest that simpler classifiers (such as QDA and its ensemble variants) may be more robust, accurate, and efficient for prostate cancer CAD problems, especially in the context of multi-site validation.

Radiomic features from pretreatment biparametric MRI predict prostate cancer biochemical recurrence: Preliminary findings.

BACKGROUND: Radiomics or computer-extracted texture features derived from MRI have been shown to help quantitatively characterize prostate cancer (PCa). Radiomics have not been explored depth in the context of predicting biochemical recurrence (BCR) of PCa. PURPOSE: To identify a set of radiomic features derived from pretreatment biparametric MRI (bpMRI) that may be predictive of PCa BCR. STUDY TYPE: Retrospective. SUBJECTS: In all, 120 PCa patients from two institutions, I1 and I2 , partitioned into training set D1 (N = 70) from I1 and independent validation set D2 (N = 50) from I2 . All patients were followed for ≥3 years. SEQUENCE: 3T, T2 -weighted (T2 WI) and apparent diffusion coefficient (ADC) maps derived from diffusion-weighted sequences. ASSESSMENT: PCa regions of interest (ROIs) on T2 WI were annotated by two experienced radiologists. Radiomic features from bpMRI (T2 WI and ADC maps) were extracted from the ROIs. A machine-learning classifier (CBCR ) was trained with the best discriminating set of radiomic features to predict BCR (pBCR ). STATISTICAL TESTS: Wilcoxon rank-sum tests with P < 0.05 were considered statistically significant. Differences in BCR-free survival at 3 years using pBCR was assessed using the Kaplan-Meier method and compared with Gleason Score (GS), PSA, and PIRADS-v2. RESULTS: Distribution statistics of co-occurrence of local anisotropic gradient orientation (CoLlAGe) and Haralick features from T2 WI and ADC were associated with BCR (P < 0.05) on D1 . CBCR predictions resulted in a mean AUC = 0.84 on D1 and AUC = 0.73 on D2 . A significant difference in BCR-free survival between the predicted classes (BCR + and BCR-) was observed (P = 0.02) on D2 compared to those obtained from GS (P = 0.8), PSA (P = 0.93) and PIRADS-v2 (P = 0.23). DATA CONCLUSION: Radiomic features from pretreatment bpMRI can be predictive of PCa BCR after therapy and may help identify men who would benefit from adjuvant therapy. LEVEL OF EVIDENCE: 4 Technical Efficacy: Stage 5 J. Magn. Reson. Imaging 2018;48:1626-1636.