Performance of a Region of Interest-based Algorithm in Diagnosing International Society of Urological Pathology Grade Group ≥2 Prostate Cancer on the MRI-FIRST Database-CAD-FIRST Study.

BACKGROUND AND OBJECTIVE: Prostate multiparametric magnetic resonance imaging (MRI) shows high sensitivity for International Society of Urological Pathology grade group (GG) ≥2 cancers. Many artificial intelligence algorithms have shown promising results in diagnosing clinically significant prostate cancer on MRI. To assess a region-of-interest-based machine-learning algorithm aimed at characterising GG ≥2 prostate cancer on multiparametric MRI. METHODS: The lesions targeted at biopsy in the MRI-FIRST dataset were retrospectively delineated and assessed using a previously developed algorithm. The Prostate Imaging-Reporting and Data System version 2 (PI-RADSv2) score assigned prospectively before biopsy and the algorithm score calculated retrospectively in the regions of interest were compared for diagnosing GG ≥2 cancer, using the areas under the curve (AUCs), and sensitivities and specificities calculated with predefined thresholds (PIRADSv2 scores ≥3 and ≥4; algorithm scores yielding 90% sensitivity in the training database). Ten predefined biopsy strategies were assessed retrospectively. KEY FINDINGS AND LIMITATIONS: After excluding 19 patients, we analysed 232 patients imaged on 16 different scanners; 85 had GG ≥2 cancer at biopsy. At patient level, AUCs of the algorithm and PI-RADSv2 were 77% (95% confidence interval [CI]: 70-82) and 80% (CI: 74-85; p = 0.36), respectively. The algorithm's sensitivity and specificity were 86% (CI: 76-93) and 65% (CI: 54-73), respectively. PI-RADSv2 sensitivities and specificities were 95% (CI: 89-100) and 38% (CI: 26-47), and 89% (CI: 79-96) and 47% (CI: 35-57) for thresholds of ≥3 and ≥4, respectively. Using the PI-RADSv2 score to trigger a biopsy would have avoided 26-34% of biopsies while missing 5-11% of GG ≥2 cancers. Combining prostate-specific antigen density, the PI-RADSv2 and algorithm's scores would have avoided 44-47% of biopsies while missing 6-9% of GG ≥2 cancers. Limitations include the retrospective nature of the study and a lack of PI-RADS version 2.1 assessment. CONCLUSIONS AND CLINICAL IMPLICATIONS: The algorithm provided robust results in the multicentre multiscanner MRI-FIRST database and could help select patients for biopsy. PATIENT SUMMARY: An artificial intelligence-based algorithm aimed at diagnosing aggressive cancers on prostate magnetic resonance imaging showed results similar to expert human assessment in a prospectively acquired multicentre test database.

Characterization of high-grade prostate cancer at multiparametric MRI using a radiomic-based computer-aided diagnosis system as standalone and second reader.

PURPOSE: The purpose of this study was to develop and test across various scanners a zone-specific region-of-interest (ROI)-based computer-aided diagnosis system (CAD) aimed at characterizing, on MRI, International Society of Urological Pathology (ISUP) grade≥2 prostate cancers. MATERIALS AND METHODS: ROI-based quantitative models were selected in multi-vendor training (265 pre-prostatectomy MRIs) and pre-test (112 pre-biopsy MRIs) datasets. The best peripheral and transition zone models were combined and retrospectively assessed in internal (158 pre-biopsy MRIs) and external (104 pre-biopsy MRIs) test datasets. Two radiologists (R1/R2) retrospectively delineated the lesions targeted at biopsy in test datasets. The CAD area under the receiver operating characteristic curve (AUC) for characterizing ISUP≥2 cancers was compared to that of the Prostate Imaging-Reporting and Data System version2 (PI-RADSv2) score prospectively assigned to targeted lesions. RESULTS: The best models used the 25th apparent diffusion coefficient (ADC) percentile in transition zone and the 2nd ADC percentile and normalized wash-in rate in peripheral zone. The PI-RADSv2 AUCs were 82% (95% confidence interval [CI]: 74-87) and 86% (95% CI: 81-91) in the internal and external test datasets respectively. They were not different from the CAD AUCs obtained with R1 and R2 delineations, in the internal (82% [95% CI: 76-89], P = 0.95 and 85% [95% CI: 78-91], P = 0.55) and external (82% [95% CI: 74-91], P = 0.41 and 86% [95% CI:78-95], P = 0.98) test datasets. The CAD yielded sensitivities of 86-89% and 90-91%, and specificities of 64-65% and 69-75% in the internal and external test datasets respectively. CONCLUSION: The CAD performance for characterizing ISUP grade≥2 prostate cancers on MRI is not different from that of PI-RADSv2 score across two test datasets.

What Influences the Way Radiologists Express Themselves in Their Reports? A Quantitative Assessment Using Natural Language Processing.

Although using standardized reports is encouraged, most emergency radiological reports in France remain in free-text format that can be mined with natural language processing for epidemiological purposes, activity monitoring or data collection. These reports are obtained under various on-call conditions by radiologists with various backgrounds. Our aim was to investigate what influences the radiologists' written expressions. To do so, this retrospective multicentric study included 30,227 emergency radiological reports of computed tomography scans and magnetic resonance imaging involving exactly one body region, only with pathological findings, interpreted from 2019-09-01 to 2020-02-28 by 165 radiologists. After text pre-processing, one-word tokenization and use of dictionaries for stop words, polarity, sentiment and uncertainty, 11 variables depicting the structure and content of words and sentences in the reports were extracted and summarized to 3 principal components capturing 93.7% of the dataset variance. In multivariate analysis, the 1st principal component summarized the length and lexical diversity of the reports and was significantly influenced by the weekday, time slot, workload, number of examinations previously interpreted by the radiologist during the on-call period, type of examination, emergency level and radiologists' gender (P value range: < 0.0001-0.0029). The 2nd principal component summarized negative formulations, polarity and sentence length and was correlated with the number of examination previously interpreted by the radiologist, type of examination, emergency level, imaging modality and radiologists' experience (P value range: < 0.0001-0.0032). The last principal component summarized questioning, uncertainty and polarity and was correlated with the type of examination and emergency level (all P values < 0.0001). Thus, the length, structure and content of emergency radiological reports were significantly influenced by organizational, radiologist- and examination-related characteristics, highlighting the subjectivity and variability in the way radiologists express themselves during their clinical activity. These findings advocate for more homogeneous practices in radiological reporting and stress the need to consider these influential features when developing models based on natural language processing.

Practical clinical and radiological models to diagnose COVID-19 based on a multicentric teleradiological emergency chest CT cohort.

Our aim was to develop practical models built with simple clinical and radiological features to help diagnosing Coronavirus disease 2019 [COVID-19] in a real-life emergency cohort. To do so, 513 consecutive adult patients suspected of having COVID-19 from 15 emergency departments from 2020-03-13 to 2020-04-14 were included as long as chest CT-scans and real-time polymerase chain reaction (RT-PCR) results were available (244 [47.6%] with a positive RT-PCR). Immediately after their acquisition, the chest CTs were prospectively interpreted by on-call teleradiologists (OCTRs) and systematically reviewed within one week by another senior teleradiologist. Each OCTR reading was concluded using a 5-point scale: normal, non-infectious, infectious non-COVID-19, indeterminate and highly suspicious of COVID-19. The senior reading reported the lesions' semiology, distribution, extent and differential diagnoses. After pre-filtering clinical and radiological features through univariate Chi-2, Fisher or Student t-tests (as appropriate), multivariate stepwise logistic regression (Step-LR) and classification tree (CART) models to predict a positive RT-PCR were trained on 412 patients, validated on an independent cohort of 101 patients and compared with the OCTR performances (295 and 71 with available clinical data, respectively) through area under the receiver operating characteristics curves (AUC). Regarding models elaborated on radiological variables alone, best performances were reached with the CART model (i.e., AUC = 0.92 [versus 0.88 for OCTR], sensitivity = 0.77, specificity = 0.94) while step-LR provided the highest AUC with clinical-radiological variables (AUC = 0.93 [versus 0.86 for OCTR], sensitivity = 0.82, specificity = 0.91). Hence, these two simple models, depending on the availability of clinical data, provided high performances to diagnose positive RT-PCR and could be used by any radiologist to support, modulate and communicate their conclusion in case of COVID-19 suspicion. Practically, using clinical and radiological variables (GGO, fever, presence of fibrotic bands, presence of diffuse lesions, predominant peripheral distribution) can accurately predict RT-PCR status.

Teleradiology as a relevant indicator of the impact of COVID-19 pandemic management on emergency room activities: a nationwide worrisome survey.

OBJECTIVES: To evaluate the impact of COVID-19's lockdown on radiological examinations in emergency services. METHODS: Retrospective, multicentre analysis of radiological examinations requested, via our teleradiology network, from 2017 to 2020 during two timeframes (calendar weeks 5-8 and then 12-15). We included CT scans or MRIs performed for strokes, multiple traumas (Body-CT), cranial traumas (CTr) and acute non-traumatic abdominal pain (ANTAP). We evaluated the number and percentages of examinations performed, of those with a pathological conclusion, and of examinations involving the chest. RESULTS: Our study included 25 centres in 2017, 29 in 2018, 43 in 2019 and 59 in 2020. From 2017 to 2019, the percentages of examinations were constant, which was also true for chest CTs. Each centre's number of examinations, gender distribution and patient ages were unchanged. In 2020, examinations significantly decreased: suspected strokes decreased by 36% (1052 vs 675, p < 0.001), Body-CT by 62% (349 vs 134, p < 0.001), CTr by 52% (1853 vs 895, p < 0.001) and for ANTAP, appendicitis decreased by 38% (45 vs 90, not statistically significant (NS)) sigmoiditis by 44% (98 vs 55, NS), and renal colic by 23% (376 vs 288, NS). The number of examinations per centre decreased by 13% (185.5 vs 162.5, p < 0.001), whereas the number of examinations of the "chest" region increased by 170% (1205 vs 3766, p < 0.001). CONCLUSION: Teleradiology enabled us to monitor the impact of the COVID-19 pandemic management on emergency activities, showing a global decrease in the population's use of care.

The accuracy of teleradiologists in diagnosing COVID-19 based on a French multicentric emergency cohort.

OBJECTIVES: To evaluate the accuracy of diagnoses of COVID-19 based on chest CT as well as inter-observer agreement between teleradiologists during on-call duty and senior radiologists in suspected COVID-19 patients. MATERIALS AND METHODS: From March 13, 2020, to April 14, 2020, consecutive suspected COVID-19 adult patients who underwent both an RT-PCR test and chest CT from 15 hospitals were included in this prospective study. Chest CTs were immediately interpreted by the on-call teleradiologist and were systematically blind reviewed by a senior radiologist. Readings were categorised using a five-point scale: (1) normal; (2) non-infectious findings; (3) infectious findings but not consistent with COVID-19 infection; (4) consistent with COVID-19 infection; and (5) typical appearance of COVID-19 infection. The diagnostic accuracy of chest CT and inter-observer agreement using the kappa coefficient were evaluated over the study period. RESULTS: In total, 513 patients were enrolled, of whom 244/513 (47.6%) tested positive for RT-PCR. First readings were scored 4 or 5 in 225/244 (92%) RT-PCR+ patients, and between 1 and 3 in 201/269 (74.7%) RT-PCR- patients. The data were highly consistent (weighted kappa = 0.87) and correlated with RT-PCR (p < 0.001, AUC1st-reading = 0.89, AUC2nd-reading = 0.93). The negative predictive value for scores of 4 or 5 was 0.91-0.92, and the PPV for a score of 5 was 0.89-0.96 at the first and second readings, respectively. Diagnostic accuracy was consistent over the study period, irrespective of a variable prevalence rate. CONCLUSION: Chest CT demonstrated high diagnostic accuracy with strong inter-observer agreement between on-call teleradiologists with varying degrees of experience and senior radiologists over the study period. KEY POINTS: • The accuracy of readings by on-call teleradiologists, relative to second readings by senior radiologists, demonstrated a sensitivity of 0.75-0.79, specificity of 0.92-0.97, NPV of 0.80-0.83, and PPV of 0.89-0.96, based on "typical appearance," as predictive of RT-PCR+. • Inter-observer agreement between the first reading in the emergency setting and the second reading by the senior emergency teleradiologist was excellent (weighted kappa = 0.87).

Whole-Body CT in Patients with Multiple Traumas: Factors Leading to Missed Injury.

Purpose To determine radiologic and clinical markers predictive of missed injuries at early whole-body CT image interpretation. Materials and Methods For this retrospective study, 2354 consecutive whole-body CT examinations were performed in patients with multiple traumas from 26 hospitals interpreted at a teleradiology center study during on-call period from February 2011 to September 2016. All whole-body CT images were interpreted by the on-call radiologist and reviewed within 12-48 hours by another radiologist to detect missed injury as the standard of reference. The first and review reports of all examinations were retrospectively reviewed. Univariable and multivariable logistic regression with a stepwise selection method were performed to identify clinical and radiologic predictors of missed injury. Results This study included 639 women (27.1%) and 1715 men (72.8%). The median age of men, women, and the entire population was 34 years (age range, 1-96 years). On a per-scan basis, there were 304 (12.9%) missed injuries and 59 (2.5%) were clinically significant. On a per-injury basis, the missed injury rate was 530 of 5979 (8.8%). More than two injured body parts (odds ratio, 1.4 [95% confidence interval: 1.1, 1.8]; P = .01), patient age older than 30 years (odds ratio, 2.8 [95% confidence interval: 2.1, 3.8]; P < .001), and an initial clinical severity class of 1 (odds ratio, 1.9 [95% confidence interval: 1.3, 2.8]; P < .001) were independent predictive factors of missed injury. Conclusion Multiple traumas with more than two injured body parts, age older than 30 years, or an initial clinical severity class of 1 were associated with missed injury at whole-body CT. © RSNA, 2018 Online supplemental material is available for this article. See also the editorial by Novelline in this issue.

Characterization of Prostate Cancer with Gleason Score of at Least 7 by Using Quantitative Multiparametric MR Imaging: Validation of a Computer-aided Diagnosis System in Patients Referred for Prostate Biopsy.

Purpose To determine the performance of a computer-aided diagnosis (CAD) system trained at characterizing cancers in the peripheral zone (PZ) with a Gleason score of at least 7 in patients referred for multiparametric magnetic resonance (MR) imaging before prostate biopsy. Materials and Methods Two institutional review board-approved prospective databases of patients who underwent multiparametric MR imaging before prostatectomy (database 1) or systematic and targeted biopsy (database 2) were retrospectively used. All patients gave informed consent for inclusion in the databases. A CAD combining the 10th percentile of the apparent diffusion coefficient and the time to peak of enhancement was trained to detect cancers in the PZ with a Gleason score of at least 7 in 106 patients from database 1. The CAD was tested in 129 different patients from database 2. All targeted lesions were prospectively scored at biopsy by using a five-level Likert score. The CAD scores were retrospectively calculated. Biopsy results were used as the reference standard. Areas under the receiver operating characteristic curves (AUCs) were computed for CAD and Likert scores by using binormal smoothing for per-lesion and per-lobe analyses, and a density function for per-patient analysis. Results The CAD outperformed the Likert score in the overall population and all subgroups, except in the transition zone. The difference was statistically significant for the overall population (AUC, 0.95 [95% confidence interval {CI}: 0.90, 0.98] vs 0.88 [95% CI: 0.68, 0.96]; P = .02) at per-patient analysis, and for less-experienced radiologists (<1 year) at per-lesion (AUC, 0.90 [95% CI: 0.81, 0.95] vs 0.83 [95% CI: 0.73, 0.90]; P = .04) and per-lobe (AUC, 0.92 [95% CI: 0.80, 0.96] vs 0.84 [95% CI: 0.72, 0.91]; P = .04) analysis. Conclusion The CAD outperformed the Likert score prospectively assigned at biopsy in characterizing cancers with a Gleason score of at least 7. © RSNA, 2018 Online supplemental material is available for this article.

Diagnostic value and relative weight of sequence-specific magnetic resonance features in characterizing clinically significant prostate cancers.

PURPOSE: To assess the diagnostic weight of sequence-specific magnetic resonance features in characterizing clinically significant prostate cancers (csPCa). MATERIALS AND METHODS: We used a prospective database of 262 patients who underwent T2-weighted, diffusion-weighted, and dynamic contrast-enhanced (DCE) imaging before prostatectomy. For each lesion, two independent readers (R1, R2) prospectively defined nine features: shape, volume (V_Max), signal abnormality on each pulse sequence, number of pulse sequences with a marked (S_Max) and non-visible (S_Min) abnormality, likelihood of extracapsular extension (ECE) and PSA density (dPSA). Overall likelihood of malignancy was assessed using a 5-level Likert score. Features were evaluated using the area under the receiver operating characteristic curve (AUC). csPCa was defined as Gleason ≥7 cancer (csPCa-A), Gleason ≥7(4+3) cancer (csPCa-B) or Gleason ≥7 cancer with histological extraprostatic extension (csPCa-C). RESULTS: For csPCa-A, the Signal1 model (S_Max+S_Min) provided the best combination of signal-related variables, for both readers. The performance was improved by adding V_Max, ECE and/or dPSA, but not shape. All models performed better with DCE findings than without. When moving from csPCa-A to csPCa-B and csPCa-C definitions, the added value of V_Max, dPSA and ECE increased as compared to signal-related variables, and the added value of DCE decreased. For R1, the best models were Signal1+ECE+dPSA (AUC = 0,805 [95%CI:0,757-0,866]), Signal1+V_Max+dPSA (AUC = 0.823 [95%CI:0.760-0.893]) and Signal1+ECE+dPSA [AUC = 0.840 (95%CI:0.774-0.907)] for csPCa-A, csPCA-B and csPCA-C respectively. The AUCs of the corresponding Likert scores were 0.844 [95%CI:0.806-0.877, p = 0.11], 0.841 [95%CI:0.799-0.876, p = 0.52]) and 0.849 [95%CI:0.811-0.884, p = 0.49], respectively. For R2, the best models were Signal1+V_Max+dPSA (AUC = 0,790 [95%CI:0,731-0,857]), Signal1+V_Max (AUC = 0.813 [95%CI:0.746-0.882]) and Signal1+ECE+V_Max (AUC = 0.843 [95%CI: 0.781-0.907]) for csPCa-A, csPCA-B and csPCA-C respectively. The AUCs of the corresponding Likert scores were 0. 829 [95%CI:0.791-0.868, p = 0.13], 0.790 [95%CI:0.742-0.841, p = 0.12]) and 0.808 [95%CI:0.764-0.845, p = 0.006]), respectively. CONCLUSION: Combination of simple variables can match the Likert score's results. The optimal combination depends on the definition of csPCa.

Stiffness of benign and malignant prostate tissue measured by shear-wave elastography: a preliminary study.

OBJECTIVES: To measure benign and malignant prostate tissue stiffness using shear-wave elastography (SWE). METHODS: Thirty consecutive patients underwent transrectal SWE in the axial and sagittal planes before prostatectomy. After reviewing prostatectomy specimens, two radiologists measured stiffness in regions corresponding to cancers, lateral and median benign peripheral zone (PZ) and benign transition zone (TZ). RESULTS: Cancers were stiffer than benign PZ and TZ. All tissue classes were stiffer on sagittal than on axial imaging, in TZ than in PZ, and in median PZ than in lateral PZ. At multivariate analysis, the nature of tissue (benign or malignant; P < 0.00001), the imaging plane (axial or sagittal; P < 0.00001) and the location within the prostate (TZ, median PZ or lateral PZ; P = 0.0065) significantly and independently influenced tissue stiffness. On axial images, the thresholds maximising the Youden index in TZ, lateral PZ and median PZ were respectively 62 kPa, 33 kPa and 49 kPa. On sagittal images, the thresholds were 76 kPa, 50 kPa and 72 kPa, respectively. CONCLUSIONS: SWE can distinguish prostate malignant and benign tissues. Tissue stiffness is influenced by the imaging plane and the location within the gland. KEY POINTS: • Prostate cancers were stiffer than the benign peripheral zone • All tissue classes were stiffer on sagittal than on axial imaging • All tissue classes were stiffer in the transition zone than in the peripheral zone • All tissue classes were stiffer in the median than in the lateral peripheral zone • Taking into account imaging plane and zonal anatomy can improve cancer detection.

Can pre- and postoperative magnetic resonance imaging predict recurrence-free survival after whole-gland high-intensity focused ablation for prostate cancer?

OBJECTIVES: Our aim was to assess whether magnetic resonance imaging (MRI) features predict recurrence-free survival (RFS) after prostate cancer high-intensity focused ultrasound (HIFU) ablation. METHODS: We retrospectively selected 81 patients who underwent (i) whole-gland HIFU ablation between 2007 and 2011 as first-line therapy or salvage treatment after radiotherapy or brachytherapy, and (ii) pre- and postoperative MRI. On preoperative imaging, two senior (R1, R2) and one junior (R3) readers assessed the number of sectors invaded by the lesion with the highest Likert score (dominant lesion) using a 27-sector diagram. On postoperative imaging, readers assessed destruction of the dominant lesion using a three-level score. Multivariate analysis included the number of sectors invaded by the dominant lesion, its Likert and destruction scores, the pre-HIFU prostate-specific antigen (PSA) level, Gleason score, and the clinical setting (primary/salvage). RESULTS: The most significant predictor was the number of prostate sectors invaded by the dominant lesion for R2 and R3 (p≤0.001) and the destruction score of the dominant lesion for R1 (p = 0.011). The pre-HIFU PSA level was an independent predictor for R2 (p = 0.014), but with only marginal significance for R1 (p = 0.059) and R3 (p = 0.053). CONCLUSION: The dominant lesion's size and destruction assessed by MRI provide independent prognostic information compared with usual predictors. KEY POINTS: • The size of the MR-dominant lesion significantly influences post-HIFU recurrence-free survival. • The destruction score of the MR-dominant lesion predicts post-HIFU recurrence-free survival. • Patients with a completely devascularized MR-dominant lesion show better recurrence-free survival • Pre- and post-HIFU MRI provide prognostic information independent of usual predictors.

Accuracy of Elastic Fusion of Prostate Magnetic Resonance and Transrectal Ultrasound Images under Routine Conditions: A Prospective Multi-Operator Study.

PURPOSE: To evaluate in unselected patients imaged under routine conditions the co-registration accuracy of elastic fusion between magnetic resonance (MR) and ultrasound (US) images obtained by the Koelis Urostation™. MATERIALS AND METHODS: We prospectively included 15 consecutive patients referred for placement of intraprostatic fiducials before radiotherapy and who gave written informed consent by signing the Institutional Review Board-approved forms. Three fiducials were placed in the prostate under US guidance in standardized positions (right apex, left mid-gland, right base) using the Koelis Urostation™. Patients then underwent prostate MR imaging. Four operators outlined the prostate on MR and US images and an elastic fusion was retrospectively performed. Fiducials were used to measure the overall target registration error (TRE3D), the error along the antero-posterior (TREAP), right-left (TRERL) and head-feet (TREHF) directions, and within the plane orthogonal to the virtual biopsy track (TRE2D). RESULTS: Median TRE3D and TRE2D were 3.8-5.6 mm, and 2.5-3.6 mm, respectively. TRE3D was significantly influenced by the operator (p = 0.013), fiducial location (p = 0.001) and 3D axis orientation (p<0.0001). The worst results were obtained by the least experienced operator. TRE3D was smaller in mid-gland and base than in apex (average difference: -1.21 mm (95% confidence interval (95%CI): -2.03; -0.4) and -1.56 mm (95%CI: -2.44; -0.69) respectively). TREAP and TREHF were larger than TRERL (average difference: +1.29 mm (95%CI: +0.87; +1.71) and +0.59 mm (95%CI: +0.1; +0.95) respectively). CONCLUSIONS: Registration error values were reasonable for clinical practice. The co-registration accuracy was significantly influenced by the operator's experience, and significantly poorer in the antero-posterior direction and at the apex.

Quantitative Analysis of Prostate Multiparametric MR Images for Detection of Aggressive Prostate Cancer in the Peripheral Zone: A Multiple Imager Study.

Purpose To assess the intermanufacturer variability of quantitative models in discriminating cancers with a Gleason score of at least 7 among peripheral zone (PZ) lesions seen at 3-T multiparametric magnetic resonance (MR) imaging. Materials and Methods An institutional review board-approved prospective database of 257 patients who gave written consent and underwent T2-weighted, diffusion-weighted, and dynamic contrast material-enhanced imaging before prostatectomy was retrospectively reviewed. It contained outlined lesions found to be suspicious for malignancy by two independent radiologists and classified as malignant or benign after correlation with prostatectomy whole-mount specimens. One hundred six patients who underwent imaging with 3-T MR systems from two manufacturers were selected (data set A, n = 72; data set B, n = 34). Eleven parameters were calculated in PZ lesions: normalized T2-weighted signal intensity, skewness and kurtosis of T2-weighted signal intensity, T2 value, wash-in rate, washout rate, time to peak (TTP), mean apparent diffusion coefficient (ADC), 10th percentile of the ADC, and skewness and kurtosis of the histogram of the ADC values. Parameters were selected on the basis of their specificity for a sensitivity of 0.95 in diagnosing cancers with a Gleason score of at least 7, and the area under the receiver operating characteristic curve (AUC) for the models was calculated. Results The model of the 10th percentile of the ADC with TTP yielded the highest AUC in both data sets. In data set A, the AUC was 0.90 (95% confidence interval [CI]: 0.85, 0.95) or 0.89 (95% CI: 0.82, 0.94) when it was trained in data set A or B, respectively. In data set B, the AUC was 0.84 (95% CI: 0.74, 0.94) or 0.86 (95% CI: 0.76, 0.95) when it was trained in data set A or B, respectively. No third variable added significantly independent information in any data set. Conclusion The model of the 10th percentile of the ADC with TTP yielded accurate results in discriminating cancers with a Gleason score of at least 7 among PZ lesions at 3 T in data from two manufacturers. (©) RSNA, 2016 Online supplemental material is available for this article.

How accurate is multiparametric MR imaging in evaluation of prostate cancer volume?

PURPOSE: To assess the factors influencing multiparametric (MP) magnetic resonance (MR) imaging accuracy in estimating prostate cancer histologic volume (Vh). MATERIALS AND METHODS: A prospective database of 202 patients who underwent MP MR imaging before radical prostatectomy was retrospectively used. Institutional review board approval and informed consent were obtained. Two independent radiologists delineated areas suspicious for cancer on images (T2-weighted, diffusion-weighted, dynamic contrast material-enhanced [DCE] pulse sequences) and scored their degree of suspicion of malignancy by using a five-level Likert score. One pathologist delineated cancers on whole-mount prostatectomy sections and calculated their volume by using digitized planimetry. Volumes of MR true-positive lesions were measured on T2-weighted images (VT2), on ADC maps (VADC), and on DCE images [VDCE]). VT2, VADC, VDCE and the greatest volume determined on images from any of the individual MR pulse sequences (Vmax) were compared with Vh (Bland-Altman analysis). Factors influencing MP MR imaging accuracy, or A, calculated as A = Vmax/Vh, were evaluated using generalized linear mixed models. RESULTS: For both readers, Vh was significantly underestimated with VT2 (P < .0001, both), VADC (P < .0001, both), and VDCE (P = .02 and P = .003, readers 1 and 2, respectively), but not with Vmax (P = .13 and P = .21, readers 1 and 2, respectively). Mean, 25th percentile, and 75th percentile, respectively, for Vmax accuracy were 0.92, 0.54, and 1.85 for reader 1 and 0.95, 0.57, and 1.77 for reader 2. At generalized linear mixed (multivariate) analysis, tumor Likert score (P < .0001), Gleason score (P = .009), and Vh (P < .0001) significantly influenced Vmax accuracy (both readers). This accuracy was good in tumors with a Gleason score of 7 or higher or a Likert score of 5, with a tendency toward underestimation of Vh; accuracy was poor in small (<0.5 cc) or low-grade (Gleason score ≤6) tumors, with a tendency toward overestimation of Vh. CONCLUSION: Vh can be estimated by using Vmax in aggressive tumors or in tumors with high Likert scores.

Characterization of prostate lesions as benign or malignant at multiparametric MR imaging: comparison of three scoring systems in patients treated with radical prostatectomy.

PURPOSE: To compare the subjective Likert score to the Prostate Imaging Reporting and Data System (PIRADS) and morphology-location-signal intensity (MLS) scores for categorization of prostate lesions as benign or malignant at multiparametric magnetic resonance (MR) imaging. MATERIALS AND METHODS: Two hundred fifteen patients who underwent T2-weighted, diffusion-weighted, and dynamic contrast material-enhanced multiparametric MR imaging of the prostate before radical prostatectomy were included in a prospective database after they signed the institutional review board-approved forms. Senior readers 1 and 2 prospectively noted the location, shape, and signal intensity of lesions on MR images from individual pulse sequences and scored each for likelihood of malignancy by using a Likert scale (range, 1-5). A junior reader (reader 3) retrospectively reviewed the database and did the same analysis. The MLS score (range, 1-13) was computed by using the readers' descriptions of the lesions. Then, the three readers again scored the lesions they described by using the PIRADS score (range, 3-15). MLS and PIRADS scores were compared with the Likert score by using their areas under the receiver operating characteristic curves. RESULTS: Areas under the receiver operating characteristic curves of the Likert, MLS, and PIRADS scores were 0.81, 0.77 (P = .03), and 0.75 (P = .01) for reader 1; 0.88, 0.74 (P < .0001), and 0.76 (P < .0001) for reader 2; and 0.81, 0.78 (P = .23), and 0.75 (P = .01) for reader 3. For diagnosing cancers with Gleason scores greater than or equal to 7, the Likert score was significantly more accurate than the others, except for the MLS score for reader 3. Weighted κ values were 0.470-0.524, 0.405-0.430, and 0.378-0.441 for the Likert, MLS, and PIRADS scores, respectively. CONCLUSION: The Likert score allowed significantly more accurate categorization of prostate lesions on MR images than did the MLS and PIRADS scores.

Prostate focal peripheral zone lesions: characterization at multiparametric MR imaging--influence of a computer-aided diagnosis system.

PURPOSE: To assess the impact of a computer-aided diagnosis (CAD) system in the characterization of focal prostate lesions at multiparametric magnetic resonance (MR) imaging. MATERIALS AND METHODS: Formal institutional review board approval was not required. Thirty consecutive 1.5-T multiparametric MR imaging studies (with T2-weighted, diffusion-weighted, and dynamic contrast material-enhanced imaging) obtained before radical prostatectomy in patients between September 2008 and February 2010 were reviewed. Twelve readers assessed the likelihood of malignancy of 88 predefined peripheral zone lesions by using a five-level (level, 0-4) subjective score (SS) in reading session 1. This was repeated 5 weeks later in reading session 2. The CAD results were then disclosed, and in reading session 3, the readers could amend the scores assigned during reading session 2. Diagnostic accuracy was assessed by using a receiver operating characteristic (ROC) regression model and was quantified with the area under the ROC curve (AUC). RESULTS: Mean AUCs were significantly lower for less experienced (<1 year) readers (P < .02 for all sessions). Seven readers improved their performance between reading sessions 1 and 2, and 12 readers improved their performance between sessions 2 and 3. The mean AUCs for reading session 1 (83.0%; 95% confidence interval [CI]: 77.9%, 88.0%) and reading session 2 (84.1%; 95% CI: 78.1%, 88.7%) were not significantly different (P = .76). Although the mean AUC for reading session 3 (87.2%; 95% CI: 81.0%, 92.0%) was higher than that for session 2, the difference was not significant (P = .08). For an SS positivity threshold of 3, the specificity of reading session 2 (79.0%; 95% CI: 71.1%, 86.4%) was not significantly different from that of session 1 (78.7%; 95% CI: 70.5%, 86.8%) but was significantly lower than that of session 3 (86.2%; 95% CI: 77.1%, 93.1%; P < .03). The sensitivity of reading session 2 (68.4%; 95% CI: 57.5%, 77.7%) was significantly higher than that of session 1 (64.0%; 95% CI: 52.9%, 73.9%; P = .003) but was not significantly different from that of session 3 (71.4%; 95% CI: 58.3%, 82.7%). CONCLUSION: A CAD system may improve the characterization of prostate lesions at multiparametric MR imaging by increasing reading specificity.

Influence of imaging and histological factors on prostate cancer detection and localisation on multiparametric MRI: a prospective study.

OBJECTIVES: To assess factors influencing prostate cancer detection on multiparametric (T2-weighted, diffusion-weighted, and dynamic contrast-enhanced) MRI. METHODS: One hundred and seventy-five patients who underwent radical prostatectomy were included. Pre-operative MRI performed at 1.5 T (n = 71) or 3 T (n = 104), with (n = 58) or without (n = 117) an endorectal coil were independently interpreted by two radiologists. A five-point subjective suspicion score (SSS) was assigned to all focal abnormalities (FAs). MR findings were then compared with whole-mount sections. RESULTS: Readers identified 192-214/362 cancers, with 130-155 false positives. Detection rates for tumours of <0.5 cc (cm(3)), 0.5-2 cc and >2 cc were 33-45/155 (21-29 %), 15-19/35 (43-54 %) and 8-9/12 (67-75 %) for Gleason ≤6, 17/27 (63 %), 42-45/51 (82-88 %) and 34/35 (97 %) for Gleason 7 and 4/5 (80 %), 13/14 (93 %) and 28/28 (100 %) for Gleason ≥8 cancers respectively. At multivariate analysis, detection rates were influenced by tumour Gleason score, histological volume, histological architecture and location (P < 0.0001), but neither by field strength nor coils used for imaging. The SSS was a significant predictor of both malignancy of FAs (P < 0.005) and aggressiveness of tumours (P < 0.00001). CONCLUSIONS: Detection rates were significantly influenced by tumour characteristics, but neither by field strength nor coils used for imaging. The SSS significantly stratified the risk of malignancy of FAs and aggressiveness of detected tumours. KEY POINTS: • Prostate cancer volume, Gleason score, architecture and location are MRI predictors of detection. • Field strength and coils used do not influence the tumour detection rate. • Multiparametric MRI is accurate for detecting aggressive tumours. • A subjective suspicion score can stratify the risk of malignancy and tumour aggressiveness.

Computer-aided diagnosis of prostate cancer in the peripheral zone using multiparametric MRI.

This study evaluated a computer-assisted diagnosis (CADx) system for determining a likelihood measure of prostate cancer presence in the peripheral zone (PZ) based on multiparametric magnetic resonance (MR) imaging, including T2-weighted, diffusion-weighted and dynamic contrast-enhanced MRI at 1.5 T. Based on a feature set derived from grey-level images, including first-order statistics, Haralick features, gradient features, semi-quantitative and quantitative (pharmacokinetic modelling) dynamic parameters, four kinds of classifiers were trained and compared: nonlinear support vector machine (SVM), linear discriminant analysis, k-nearest neighbours and naïve Bayes classifiers. A set of feature selection methods based on t-test, mutual information and minimum-redundancy-maximum-relevancy criteria were also compared. The aim was to discriminate between the relevant features as well as to create an efficient classifier using these features. The diagnostic performances of these different CADx schemes were evaluated based on a receiver operating characteristic (ROC) curve analysis. The evaluation database consisted of 30 sets of multiparametric MR images acquired from radical prostatectomy patients. Using histologic sections as the gold standard, both cancer and nonmalignant (but suspicious) tissues were annotated in consensus on all MR images by two radiologists, a histopathologist and a researcher. Benign tissue regions of interest (ROIs) were also delineated in the remaining prostate PZ. This resulted in a series of 42 cancer ROIs, 49 benign but suspicious ROIs and 124 nonsuspicious benign ROIs. From the outputs of all evaluated feature selection methods on the test bench, a restrictive set of about 15 highly informative features coming from all MR sequences was discriminated, thus confirming the validity of the multiparametric approach. Quantitative evaluation of the diagnostic performance yielded a maximal area under the ROC curve (AUC) of 0.89 (0.81-0.94) for the discrimination of the malignant versus nonmalignant tissues and 0.82 (0.73-0.90) for the discrimination of the malignant versus suspicious tissues when combining the t-test feature selection approach with a SVM classifier. A preliminary comparison showed that the optimal CADx scheme mimicked, in terms of AUC, the human experts in differentiating malignant from suspicious tissues, thus demonstrating its potential for assisting cancer identification in the PZ.