Volumetry of the dominant intraprostatic tumour lesion: intersequence and interobserver differences on multiparametric MRI

The predictive value of machine learning and nomograms for lymph node metastasis of prostate cancer: a systematic review and meta-analysis

BACKGROUND

In clinical practice, there are currently a variety of nomograms for predicting lymph node metastasis (LNM) of prostate cancer. At the same time, some scholars have introduced machine learning (ML) into the prediction of LNM of prostate cancer. However, the predictive value of nomograms and ML remains controversial. Based on this situation, this systematic review and meta-analysis was performed to explore the predictive value of various nomograms currently recommended and newly-developed ML models for LNM in prostate cancer patients.

EVIDENCE ACQUISITION

Cochrane, PubMed, Embase, and Web of Science were searched up to November 1, 2022. The risk of bias in the included studies was evaluated using the Prediction model Risk of Bias Assessment Tool (PROBAST). The concordance index (C-index), sensitivity, and specificity were adopted to evaluate the predictive accuracy of the models.

RESULTS

Thirty-one studies (18,803 patients) were included. Seven kinds of nomograms currently recommended, dominated by Briganti nomogram or MSKCC nomogram, were covered in the included studies. For newly-developed ML models, the C-index for LNM prediction in the training set and validation set was 0.846 [95%CI (0.818, 0.873)] and 0.862 [95%CI (0.819-0.905)] respectively. Most ML models in the training set were based on Logistic Regression (LR), which had a sensitivity of 0.78 [95%CI (0.70, 0.85)] and a specificity of 0.85 [95%CI (0.77, 0.90)] in the training set, and a sensitivity of 0.81 [95%CI (0.67, 0.89)] and a specificity of 0.82 [95%CI (0.75, 0.88)] in the validation set. For the recommended nomograms, the C-index in the validation set was 0.745 [95%CI (0.701, 0.790)] for the Briganti nomogram and 0.714 [95%CI (0.662, 0.765)] for the MSKCC nomogram.

CONCLUSION

The predictive accuracy of ML is superior to existing clinically recommended nomograms, and appropriate updates can be conducted to existing nomograms according to special situations.

Deep Learning Reconstruction of Diffusion-weighted MRI Improves Image Quality for Prostatic Imaging

Background Deep learning reconstruction (DLR) may improve image quality. However, its impact on diffusion-weighted imaging (DWI) of the prostate has yet to be assessed. Purpose To determine whether DLR can improve image quality of diffusion-weighted MRI at b values ranging from 1000 sec/mm² to 5000 sec/mm² in patients with prostate cancer. Materials and Methods In this retrospective study, images of the prostate obtained at DWI with a b value of 0 sec/mm², DWI with a b value of 1000 sec/mm² (DWI₁₀₀₀), DWI with a b value of 3000 sec/mm² (DWI₃₀₀₀), and DWI with a b value of 5000 sec/mm² (DWI₅₀₀₀) from consecutive patients with biopsy-proven cancer from January to June 2020 were reconstructed with and without DLR. Image quality was assessed using signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) from region-of-interest analysis and qualitatively assessed using a five-point visual scoring system (1 [very poor] to 5 [excellent]) for each high-b-value DWI sequence with and without DLR. The SNR, CNR, and visual score for DWI with and without DLR were compared with the paired t test and the Wilcoxon signed rank test with Bonferroni correction, respectively. Apparent diffusion coefficients (ADCs) from DWI with and without DLR were also compared with the paired t test with Bonferroni correction. Results A total of 60 patients (mean age, 67 years; age range, 49-79 years) were analyzed. DWI with DLR showed significantly higher SNRs and CNRs than DWI without DLR (P < .001); for example, with DWI₁₀₀₀ the mean SNR was 38.7 ± 0.6 versus 17.8 ± 0.6, respectively (P < .001), and the mean CNR was 18.4 ± 5.6 versus 7.4 ± 5.6, respectively (P < .001). DWI with DLR also demonstrated higher qualitative image quality than DWI without DLR (mean score: 4.8 ± 0.4 vs 4.0 ± 0.7, respectively, with DWI₁₀₀₀ [P = .001], 3.8 ± 0.7 vs 3.0 ± 0.8 with DWI₃₀₀₀ [P = .002], and 3.1 ± 0.8 vs 2.0 ± 0.9 with DWI₅₀₀₀ [P < .001]). ADCs derived with and without DLR did not differ substantially (P > .99). Conclusion Deep learning reconstruction improves the image quality of diffusion-weighted MRI scans of prostate cancer with no impact on apparent diffusion coefficient quantitation with a 3.0-T MRI system. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Turkbey in this issue.

Combining prostate health index and multiparametric magnetic resonance imaging in estimating the histological diameter of prostate cancer

BACKGROUND

Although multiparametric magnetic resonance imaging (mpMRI) is widely used to assess the volume of prostate cancer, it often underestimates the histological tumor boundary. The aim of this study was to evaluate the feasibility of combining prostate health index (PHI) and mpMRI to estimate the histological tumor diameter and determine the safety margin during treatment of prostate cancer.

METHODS

We retrospectively enrolled 72 prostate cancer patients who underwent radical prostatectomy and had received PHI tests and mpMRI before surgery. We compared the discrepancy between histological and radiological tumor diameter stratified by Prostate Imaging-Reporting and Data System (PI-RADS) score, and then assessed the influence of PHI on the discrepancy between low PI-RADS (2 or 3) and high PI-RADS (4 or 5) groups.

RESULTS

The mean radiological and histological tumor diameters were 1.60 cm and 2.13 cm, respectively. The median discrepancy between radiological and histological tumor diameter of PI-RADS 4 or 5 lesions was significantly greater than that of PI-RADS 2 or 3 lesions (0.50 cm, IQR (0.00-0.90) vs. 0.00 cm, IQR (-0.10-0.20), p = 0.02). In the low PI-RADS group, the upper limit of the discrepancy was 0.2 cm; so the safety margin could be set at 0.1 cm. In the high PI-RADS group, the upper limits of the discrepancy were 1.2, 1.6, and 2.2 cm in men with PHI < 30, 30-60, and > 60; so the safety margin could be set at 0.6, 0.8, and 1.1 cm, respectively.

CONCLUSIONS

Radiological tumor diameter on mpMRI often underestimated the histological tumor diameter, especially for PI-RADS 4 or 5 lesions. Combining mpMRI and PHI may help to better estimate the histological tumor diameter.

Analysis of outcomes after non-contour-based dose painting of dominant intra-epithelial lesion in intra-operative low-dose rate brachytherapy

Purpose

To compare the outcomes of patients with intermediate risk prostate cancer (IR-PCa) treated with low-dose rate I-125 seed brachytherapy (LDR-BT) and targeted dose painting of a histologic dominant intra-epithelial lesion (DIL) to those without a DIL.

Methods

455 patients with IR-PCa were treated at a single center with intra-operatively planned LDR-BT, each following the same in-house dose constraints. Patients with a DIL on pathology had hot spots localized to that region but no specific contouring during the procedure.

Results

396 (87%) patients had a DIL. Baseline tumor characteristics and overall prostate dosimetry were similar between patients with and without DIL except the median number of biopsy cores taken: 10 (10–12) vs 12 (10–12) (p = 0.002).

19 (5%) and 18 (5%) of patients with and 1 (2%) and 0 (0%) of those without DIL experienced CTCAE grade 2 and 3 toxicity respectively. Overall, toxicity grade did not significantly correlate with presence of DIL (p = 0.10).

Estimated 7-year freedom from biochemical failure (FFBF) was 84% (95% confidence interval: 79–89) and 70% (54–89) in patients with and without a DIL (log-rank p = 0.315). In DIL patients, cox regression revealed location of DIL (“Base” vs “Apex” HR: 1.03; 1.00–1.06; p = 0.03) and older age (70 vs 60 HR: 1.62; 1.06–2.49; p = 0.03) was associated with poor FFBF.

Conclusions

Targeting DIL through dose painting during intraoperatively planned LDR-BT provided no statistically significant change in FFBF. Patients with DILs in the prostate base had slightly lower FFBF despite DIL boost.

How accurately can multiparametric magnetic resonance imaging measure the tumour volume of a prostate cancer? Results of a systematic review

The tumour volume of a cancer within the prostate gland is commonly measured with multiparametric MRI. The measurement has a role in many clinical scenarios including focal therapy, but the accuracy of it has never been systematically reviewed. We included articles if they compared tumour volume measurements obtained by mpMRI with a reference volume measurement obtained after radical prostatectomy. Correlation and concordance statistics were summarised. A simple accuracy score was derived by dividing the given mean or median mpMRI volume by the histopathological reference volume. Factors affecting the accuracy were noted. Scores for potential bias and quality were calculated for each article. A total of 18 articles describing 1438 patients were identified. Nine articles gave Pearson's correlation scores, with a median value of 0.75 but the range was wide (0.42-0.97). A total of 11 articles reported mean values for volume while 9 reported median values. For all 18 articles, the mean or median values for MRI volumes were lower than the corresponding reference values suggesting consistent underestimation. For articles reporting mean and median values for volume, the median accuracy scores were 0.83 and 0.80, respectively. The accuracy was higher for tumours of greater volume, higher grade and when an endorectal coil was used. Accuracy did not seem to improve over time, with a 3 Tesla magnet or by applying a shrinkage factor to the reference measurement. Most studies showed evidence of at least moderate bias, and their quality was highly variable, but neither of these appeared to affect accuracy.

Utility of index lesion volume assessed by multiparametric MRI combined with Gleason grade for assessment of lymph node involvement in patients with high-risk prostate cancer

PURPOSE

We examined the potential predictors of lymph node involvement and evaluated whether index lesion volume assessed using multiparametric magnetic resonance imaging is associated with lymph node involvement among patients with high-risk prostate cancer.

METHODS

Extended pelvic lymph node dissection was used to evaluate patients with lymph node involvement. We retrospectively analyzed consecutive 102 patients with high-risk prostate cancer who underwent extended pelvic lymph node dissection at our institution between 2011 and 2017. To evaluate the index lesion volume at multiparametric magnetic resonance imaging (mrV), lesions were manually contoured on each T2-weighted axial slice in combination with diffusion-weighted and dynamic contrast-enhanced magnetic resonance imaging and integrated using image analysis software. Logistic regression analysis was performed to identify predictors of lymph node involvement.

RESULTS

The median mrV was 1.4 ml (range 0-30.1 ml), and the median number of resected lymph nodes was 14 (range 7-38). Among 102 patients, 28 (28%) had lymph node involvement. Multivariate analysis identified significant predictors of lymph node involvement as follows: biopsy Gleason-grade group 5 (odds ratio = 17.2; 95% confidence interval, 2.1-299.0; P = 0.005), preoperative mrV (odds ratio = 1.14; 95% confidence interval, 1.02-1.30; P = 0.025) and percentage of positive cores with highest Gleason-grade group (odds ratio = 1.05; 95% confidence interval, 1.01-1.10; P = 0.005). Lymph node involvement was prevalent (69%) among tumors with Gleason-grade group 5 and mrV ≥3.4 ml, but was infrequently (10%) present among tumors with Gleason-grade group ≤4 and mrV <3.4 ml.

CONCLUSIONS

The combination of biopsy Gleason-grade and mrV may serve as a useful tool to stratify patients according to their risk of nodal metastases.

Prognostic Value of Pretreatment MRI in Patients With Prostate Cancer Treated With Radiation Therapy: A Systematic Review and Meta-Analysis

OBJECTIVE. Despite a substantial increase in the use of MRI for pretreatment evaluation of prostate cancer, its prognostic value in patients undergoing radiation therapy (RT) is not well known. Therefore, the purpose of this study was to systematically review the literature and perform a meta-analysis on the prognostic value of pretreatment MRI in patients with prostate cancer who underwent external beam radiation therapy (EBRT) or brachytherapy. MATERIALS AND METHODS. PubMed and Embase databases were searched for studies published on or before March 13, 2019. We included studies that evaluated pretreatment MRI as a prognostic factor in prostate cancer regarding biochemical recurrence (BCR), metastatic failure, and overall or cancer-specific mortality. Effect sizes were measured in terms of the hazard ratio (HR) and were meta-analytically pooled using the random-effects model. The quality of the studies was independently evaluated using the Quality in Prognostic Studies tool. RESULTS. Twelve studies (2205 patients) were included. All studies assessed BCR; metastasis was evaluated in three studies, and mortality was evaluated in one study. Extraprostatic extension (EPE), seminal vesicle invasion (SVI), large tumor size or volume, number of sextants involved, and tumor involvement of prostatic apex were significant prognostic factors of BCR (pooled HRs = 1.50-4.47). EPE, larger tumor size, greater tumor volume, presence of metastatic pelvic lymph nodes (LNs), and presence of SVI were significant risk factors for metastasis (pooled HRs = 1.12-11.96). Pelvic LN metastasis was significantly predictive of cancer-specific mortality (HR = 4.45 [95% CI, 1.30-15.23]). CONCLUSION. Several pretreatment MRI findings were significant prognostic factors in patients with prostate cancer who underwent RT.

The natural history of prostate cancer on MRI: lessons from an active surveillance cohort

BACKGROUND

Multiparametric magnetic resonance imaging (mpMRI) is being used increasingly in the setting of active surveillance (AS) for prostate cancer. We investigated changes in the mpMRI appearance of lesions on AS, to show the variability of volume measurements in visible lesions and assess change in lesion size according to grade.

METHODS

We retrospectively retrieved 86 men on AS (NICE guidelines) with more than one mpMRI (the first before 2013). Two radiologists, in consensus, were blinded to patient demographics and date of scan. The scans were randomly reported to reduce any bias. For visible lesions, we measured volume by planimetry on the sequence best showing the most conspicuous (index) tumour and attributed a 5-point Likert score.

RESULTS

43/86 men did not have a visible lesion on the initial mpMRI (≤2/5). Of these, 5/43 had developed a lesion scoring ≥3/5 at a median of 3.6 years of follow up. 40/86 had a lesion scoring ≥3/5 on two or more scans. There was a significant increase in volume over 3.6 years by a median of 10% (p < 0.01)-by a median of 6% for Gleason 3+3 and 18% for 3+4 (p = 0.058). Thirty-five men had a visible lesion on two scans separated by <2 years; of these, 21/35 showed a 78% median increase in tumour size between the two scans and 11/35 showed an apparent 25% median decrease in lesion size.

CONCLUSIONS

A total of 17% of men with no visible lesion developed a visible lesion at a median follow up of 3.6 years. It is possible to show significant growth in patients with a visible lesion, but variability in volume measurements between scans means that it is difficult to reliably detect increases of this order. This variability may inform the design of mpMRI protocols in AS and the time between follow up scans.

MRI-guided localization of the dominant intraprostatic lesion and dose analysis of volumetric modulated arc therapy planning for prostate cancer

PURPOSE

Primary radiation therapy is a curative treatment option for prostate cancer. The aim of this study was to evaluate the detection of the dominant intraprostatic lesion (DIL) with magnetic resonance imaging (MRI) for radiotherapy treatment planning, the comparison with transrectal ultrasound (TRUS)-guided biopsies and the examination of the dose distribution in relation to the DIL location.

MATERIALS AND METHODS

In all, 54 patients with treatment planning MRI for primary radiotherapy of prostate cancer from 03/2015 to 03/2017 at the Universitätsklinikum Würzburg were identified. The localization of the DIL was based on MRI with T2- and diffusion-weighted imaging. After registration of the MR image sets within Pinnacle³ (Philips Radiation Oncology Systems, Fitchburg, WI, USA), the dose distribution was analyzed. The location of the DIL was compared to the pathology reports in a side-based manner.

RESULTS

The DIL mean dose (Dmean) was 77.51 ± 0.77 Gy and in 50/51 cases within the tolerance range or exceeded the prescribed dose. There was a significant difference in Dmean between ventral (n = 21) and dorsal (n = 30) DIL (77.87 ± 0.67 vs. 77.26 ± 0.77 Gy; p = 0.005). MRI-guided localization showed an accuracy and sensitivity of up to 78.8% and 82.1% for inclusion of secondary lesions, respectively.

CONCLUSION

Up to 82.1% of histologically verified intraprostatic lesions were identified in the context of MRI-guided radiotherapy treatment planning. As expected, dorsal DIL tend to be minimally underdosed in comparison to ventral DIL. Adequate dose coverage was achieved in over 98% of patients.

Monitoring Tumor Volume in Patients With Prostate Cancer Undergoing Active Surveillance: Is MRI Apparent Diffusion Coefficient Indicative of Tumor Growth?

OBJECTIVE

The purpose of this study was to measure longitudinal change in tumor volume of the dominant intraprostatic lesion and determine whether baseline apparent diffusion coefficient (ADC) and change in ADC are indicative of tumor growth in patients with prostate cancer undergoing active surveillance.

SUBJECTS AND METHODS

The study group included 151 men (mean age, 68.1 ± 7.4 [SD] years; range, 50-83 years) undergoing active surveillance with 3D whole prostate, zonal, and tumor volumetric findings documented at endorectal MRI examinations performed at two time points (median interval, 1.9 years). Tumor (location confirmed at transrectal ultrasound or template biopsy) ADC was measured on the slice with the largest lesion. Twenty randomly selected patients had the measurements repeated by the same observer after a greater than 4-month interval, and the limits of agreement of measurements were calculated. Tumor volume increases greater than the upper limit of agreement were designated measurable growth, and their baseline ADCs and change in ADC were compared with those of tumors without measurable growth (independent-samples t test).

RESULTS

Fifty-two (34.4%) tumors increased measurably in volume. Baseline ADC and tumor volume were negatively correlated (r = -0.42, p = 0.001). Baseline ADC values did not differ between those with and those without measurable growth (p = 0.06), but change in ADC was significantly different (-6.8% ± 12.3% for those with measurable growth vs 0.23% ± 10.1% for those without, p = 0.0005). Percentage change in tumor volume and percentage change in ADC were negatively correlated (r = -0.31, p = 0.0001). A 5.8% reduction in ADC indicated a measurable increase in tumor volume with 54.9% sensitivity and 77.0% specificity (AUC, 0.67).

CONCLUSION

Tumor volume increased measurably in 34.4% of men after 2 years of active surveillance. Change in ADC may be used to identify tumors with measurable growth.

Assessment of Geometric Distortion in Six Clinical Scanners Using a 3D-Printed Grid Phantom

A cost-effective regularly structured three-dimensional (3D) printed grid phantom was developed to enable the quantification of machine-related magnetic resonance (MR) distortion. This phantom contains reference features, “point-like” objects, or vertices, which resulted from the intersection of mesh edges in 3D space. 3D distortions maps were computed by comparing the locations of corresponding features in both MR and computer tomography (CT) data sets using normalized cross correlation. Results are reported for six MRI scanners at both 1.5 T and 3.0 T field strengths within our institution. Mean Euclidean distance error for all MR volumes in this study, was less than 2 mm. The maximum detected error for the six scanners ranged from 2.4 mm to 6.9 mm. The conclusions in this study agree well with previous studies that indicated that MRI is quite accurate near the centre of the field but is more spatially inaccurate toward the edges of the magnetic field.