Artificial intelligence and radiologists in prostate cancer detection on MRI (PI-CAI): an international, paired, non-inferiority, confirmatory study.

BACKGROUND: Artificial intelligence (AI) systems can potentially aid the diagnostic pathway of prostate cancer by alleviating the increasing workload, preventing overdiagnosis, and reducing the dependence on experienced radiologists. We aimed to investigate the performance of AI systems at detecting clinically significant prostate cancer on MRI in comparison with radiologists using the Prostate Imaging-Reporting and Data System version 2.1 (PI-RADS 2.1) and the standard of care in multidisciplinary routine practice at scale. METHODS: In this international, paired, non-inferiority, confirmatory study, we trained and externally validated an AI system (developed within an international consortium) for detecting Gleason grade group 2 or greater cancers using a retrospective cohort of 10 207 MRI examinations from 9129 patients. Of these examinations, 9207 cases from three centres (11 sites) based in the Netherlands were used for training and tuning, and 1000 cases from four centres (12 sites) based in the Netherlands and Norway were used for testing. In parallel, we facilitated a multireader, multicase observer study with 62 radiologists (45 centres in 20 countries; median 7 [IQR 5-10] years of experience in reading prostate MRI) using PI-RADS (2.1) on 400 paired MRI examinations from the testing cohort. Primary endpoints were the sensitivity, specificity, and the area under the receiver operating characteristic curve (AUROC) of the AI system in comparison with that of all readers using PI-RADS (2.1) and in comparison with that of the historical radiology readings made during multidisciplinary routine practice (ie, the standard of care with the aid of patient history and peer consultation). Histopathology and at least 3 years (median 5 [IQR 4-6] years) of follow-up were used to establish the reference standard. The statistical analysis plan was prespecified with a primary hypothesis of non-inferiority (considering a margin of 0·05) and a secondary hypothesis of superiority towards the AI system, if non-inferiority was confirmed. This study was registered at ClinicalTrials.gov, NCT05489341. FINDINGS: Of the 10 207 examinations included from Jan 1, 2012, through Dec 31, 2021, 2440 cases had histologically confirmed Gleason grade group 2 or greater prostate cancer. In the subset of 400 testing cases in which the AI system was compared with the radiologists participating in the reader study, the AI system showed a statistically superior and non-inferior AUROC of 0·91 (95% CI 0·87-0·94; p<0·0001), in comparison to the pool of 62 radiologists with an AUROC of 0·86 (0·83-0·89), with a lower boundary of the two-sided 95% Wald CI for the difference in AUROC of 0·02. At the mean PI-RADS 3 or greater operating point of all readers, the AI system detected 6·8% more cases with Gleason grade group 2 or greater cancers at the same specificity (57·7%, 95% CI 51·6-63·3), or 50·4% fewer false-positive results and 20·0% fewer cases with Gleason grade group 1 cancers at the same sensitivity (89·4%, 95% CI 85·3-92·9). In all 1000 testing cases where the AI system was compared with the radiology readings made during multidisciplinary practice, non-inferiority was not confirmed, as the AI system showed lower specificity (68·9% [95% CI 65·3-72·4] vs 69·0% [65·5-72·5]) at the same sensitivity (96·1%, 94·0-98·2) as the PI-RADS 3 or greater operating point. The lower boundary of the two-sided 95% Wald CI for the difference in specificity (-0·04) was greater than the non-inferiority margin (-0·05) and a p value below the significance threshold was reached (p<0·001). INTERPRETATION: An AI system was superior to radiologists using PI-RADS (2.1), on average, at detecting clinically significant prostate cancer and comparable to the standard of care. Such a system shows the potential to be a supportive tool within a primary diagnostic setting, with several associated benefits for patients and radiologists. Prospective validation is needed to test clinical applicability of this system. FUNDING: Health~Holland and EU Horizon 2020.

Reducing femoral flow artefacts in radial magnetic resonance fingerprinting of the prostate using region-optimised virtual coils.

High acceleration factors in radial magnetic resonance fingerprinting (MRF) of the prostate lead to strong streak-like artefacts from flow in the femoral blood vessels, possibly concealing important anatomical information. Region-optimised virtual (ROVir) coils is a beamforming-based framework to create virtual coils that maximise signal in a region of interest while minimising signal in a region of interference. In this study, the potential of removing femoral flow streak artefacts in prostate MRF using ROVir coils is demonstrated in silico and in vivo. The ROVir framework was applied to radial MRF k-space data in an automated pipeline designed to maximise prostate signal while minimising signal from the femoral vessels. The method was tested in 15 asymptomatic volunteers at 3 T. The presence of streaks was visually assessed and measurements of whole prostate T1, T2 and signal-to-noise ratio (SNR) with and without streak correction were examined. In addition, a purpose-built simulation framework in which blood flow through the femoral vessels can be turned on and off was used to quantitatively evaluate ROVir's ability to suppress streaks in radial prostate MRF. In vivo it was shown that removing selected ROVir coils visibly reduces streak-like artefacts from the femoral blood flow, without increasing the reconstruction time. On average, 80% of the prostate SNR was retained. A similar reduction of streaks was also observed in silico, while the quantitative accuracy of T1 and T2 mapping was retained. In conclusion, ROVir coils efficiently suppress streaking artefacts from blood flow in radial MRF of the prostate, thereby improving the visual clarity of the images, without significant sacrifices to acquisition time, reconstruction time and accuracy of quantitative values. This is expected to help enable T1 and T2 mapping of prostate cancer in clinically viable times, aiding differentiation between prostate cancer from noncancer and healthy prostate tissue.

Added Value of [18F]PSMA-1007 PET/CT and PET/MRI in Patients With Biochemically Recurrent Prostate Cancer: Impact on Detection Rates and Clinical Management.

BACKGROUND: Prostate-specific membrane antigen (PSMA) positron emission tomography (PET) can change management in a large fraction of patients with biochemically recurrent prostate cancer (BCR). PURPOSE: To investigate the added value of PET to MRI and CT for this patient group, and to explore whether the choice of the PET paired modality (PET/MRI vs. PET/CT) impacts detection rates and clinical management. STUDY TYPE: Retrospective. SUBJECTS: 41 patients with BCR (median age [range]: 68 [55-78]). FIELD STRENGTH/SEQUENCE: 3T, including T1-weighted gradient echo (GRE), T2-weighted turbo spin echo (TSE) and dynamic contrast-enhanced GRE sequences, diffusion-weighted echo-planar imaging, and a T1-weighted TSE spine sequence. In addition to MRI, [18F]PSMA-1007 PET and low-dose CT were acquired on the same day. ASSESSMENT: Images were reported using a five-point Likert scale by two teams each consisting of a radiologist and a nuclear medicine physician. The radiologist performed a reading using CT and MRI data and a joint reading between radiologist and nuclear medicine physician was performed using MRI, CT, and PET from either PET/MRI or PET/CT. Findings were presented to an oncologist to create intended treatment plans. Intrareader and interreader agreement analysis was performed. STATISTICAL TESTS: McNemar test, Cohen's κ, and intraclass correlation coefficients. A P-value <0.05 was considered significant. RESULTS: 7 patients had positive findings on MRI and CT, 22 patients on joint reading with PET/CT, and 18 patients joint reading with PET/MRI. For overall positivity, interreader agreement was poor for MR and CT (κ = 0.36) and almost perfect with addition of PET (PET/CT κ = 0.85, PET/MRI κ = 0.85). The addition of PET from PET/CT and PET/MRI changed intended treatment in 20 and 18 patients, respectively. Between joint readings, intended treatment was different for eight patients. DATA CONCLUSION: The addition of [18F]PSMA-1007 PET/MRI or PET/CT to MRI and CT may increase detection rates, could reduce interreader variability, and may change intended treatment in half of patients with BCR. LEVEL OF EVIDENCE: 3 TECHNICAL EFFICACY: Stage 3.

Performance of magnetic resonance imaging-based prostate cancer risk calculators and decision strategies in two large European medical centres.

OBJECTIVES: To compare the performance of currently available biopsy decision support tools incorporating magnetic resonance imaging (MRI) findings in predicting clinically significant prostate cancer (csPCa). PATIENTS AND METHODS: We retrospectively included men who underwent prostate MRI and subsequent targeted and/or systematic prostate biopsies in two large European centres. Available decision support tools were identified by a PubMed search. Performance was assessed by calibration, discrimination, decision curve analysis (DCA) and numbers of biopsies avoided vs csPCa cases missed, before and after recalibration, at risk thresholds of 5%-20%. RESULTS: A total of 940 men were included, 507 (54%) had csPCa. The median (interquartile range) age, prostate-specific antigen (PSA) level, and PSA density (PSAD) were 68 (63-72) years, 9 (7-15) ng/mL, and 0.20 (0.13-0.32) ng/mL2 , respectively. In all, 18 multivariable risk calculators (MRI-RCs) and dichotomous biopsy decision strategies based on MRI findings and PSAD thresholds were assessed. The Van Leeuwen model and the Rotterdam Prostate Cancer Risk Calculator (RPCRC) had the best discriminative ability (area under the receiver operating characteristic curve 0.86) of the MRI-RCs that could be assessed in the whole cohort. DCA showed the highest clinical utility for the Van Leeuwen model, followed by the RPCRC. At the 10% threshold the Van Leeuwen model would avoid 22% of biopsies, missing 1.8% of csPCa, whilst the RPCRC would avoid 20% of biopsies, missing 2.6% of csPCas. These multivariable models outperformed all dichotomous decision strategies based only on MRI-findings and PSAD. CONCLUSIONS: Even in this high-risk cohort, biopsy decision support tools would avoid many prostate biopsies, whilst missing very few csPCa cases. The Van Leeuwen model had the highest clinical utility, followed by the RPCRC. These multivariable MRI-RCs outperformed and should be favoured over decision strategies based only on MRI and PSAD.

The impact of pre-processing and disease characteristics on reproducibility of T2-weighted MRI radiomics features.

PURPOSE: To evaluate the reproducibility of radiomics features derived via different pre-processing settings from paired T2-weighted imaging (T2WI) prostate lesions acquired within a short interval, to select the setting that yields the highest number of reproducible features, and to evaluate the impact of disease characteristics (i.e., clinical variables) on features reproducibility. MATERIALS AND METHODS: A dataset of 50 patients imaged using T2WI at 2 consecutive examinations was used. The dataset was pre-processed using 48 different settings. A total of 107 radiomics features were extracted from manual delineations of 74 lesions. The inter-scan reproducibility of each feature was measured using the intra-class correlation coefficient (ICC), with ICC values > 0.75 considered good. Statistical differences were assessed using Mann-Whitney U and Kruskal-Wallis tests. RESULTS: The pre-processing parameters strongly influenced the reproducibility of radiomics features of T2WI prostate lesions. The setting that yielded the highest number of features (25 features) with high reproducibility was the relative discretization with a fixed bin number of 64, no signal intensity normalization, and outlier filtering by excluding outliers. Disease characteristics did not significantly impact the reproducibility of radiomics features. CONCLUSION: The reproducibility of T2WI radiomics features was significantly influenced by pre-processing parameters, but not by disease characteristics. The selected pre-processing setting yielded 25 reproducible features.

Pseudo-T2 mapping for normalization of T2-weighted prostate MRI.

OBJECTIVE: Signal intensity normalization is necessary to reduce heterogeneity in T2-weighted (T2W) magnetic resonance imaging (MRI) for quantitative analysis of multicenter data. AutoRef is an automated dual-reference tissue normalization method that normalizes transversal prostate T2W MRI by creating a pseudo-T2 map. The aim of this study was to evaluate the accuracy of pseudo-T2s and multicenter standardization performance for AutoRef with three pairs of reference tissues: fat/muscle (AutoRefF), femoral head/muscle (AutoRefFH) and pelvic bone/muscle (AutoRefPB). MATERIALS AND METHODS: T2s measured by multi-echo spin echo (MESE) were compared to AutoRef pseudo-T2s in the whole prostate (WP) and zones (PZ and TZ/CZ/AFS) for seven asymptomatic volunteers with a paired Wilcoxon signed-rank test. AutoRef normalization was assessed on T2W images from a multicenter evaluation set of 1186 prostate cancer patients. Performance was measured by inter-patient histogram intersections of voxel intensities in the WP before and after normalization in a selected subset of 80 cases. RESULTS: AutoRefFH pseudo-T2s best approached MESE T2s in the volunteer study, with no significant difference shown (WP: p = 0.30, TZ/CZ/AFS: p = 0.22, PZ: p = 0.69). All three AutoRef versions increased inter-patient histogram intersections in the multicenter dataset, with median histogram intersections of 0.505 (original data), 0.738 (AutoRefFH), 0.739 (AutoRefF) and 0.726 (AutoRefPB). DISCUSSION: All AutoRef versions reduced variation in the multicenter data. AutoRefFH pseudo-T2s were closest to experimentally measured T2s.

Divergent Regulation of Decidual Oxidative-Stress Response by NRF2 and KEAP1 in Preeclampsia with and without Fetal Growth Restriction.

Utero-placental development in pregnancy depends on direct maternal-fetal interaction in the uterine wall decidua. Abnormal uterine vascular remodeling preceding placental oxidative stress and placental dysfunction are associated with preeclampsia and fetal growth restriction (FGR). Oxidative stress is counteracted by antioxidants and oxidative repair mechanisms regulated by the transcription factor nuclear factor erythroid 2-related factor 2 (NRF2). We aimed to determine the decidual regulation of the oxidative-stress response by NRF2 and its negative regulator Kelch-like ECH-associated protein 1 (KEAP1) in normal pregnancies and preeclamptic pregnancies with and without FGR. Decidual tissue from 145 pregnancies at delivery was assessed for oxidative stress, non-enzymatic antioxidant capacity, cellular NRF2- and KEAP1-protein expression, and NRF2-regulated transcriptional activation. Preeclampsia combined with FGR was associated with an increased oxidative-stress level and NRF2-regulated gene expression in the decidua, while decidual NRF2- and KEAP1-protein expression was unaffected. Although preeclampsia with normal fetal growth also showed increased decidual oxidative stress, NRF2-regulated gene expression was reduced, and KEAP1-protein expression was increased in areas of high trophoblast density. The trophoblast-dependent KEAP1-protein expression in preeclampsia with normal fetal growth indicates control of decidual oxidative stress by maternal-fetal interaction and underscores the importance of discriminating between preeclampsia with and without FGR.

Automated reference tissue normalization of T2-weighted MR images of the prostate using object recognition.

OBJECTIVES: To develop and evaluate an automated method for prostate T2-weighted (T2W) image normalization using dual-reference (fat and muscle) tissue. MATERIALS AND METHODS: Transverse T2W images from the publicly available PROMISE12 (N = 80) and PROSTATEx (N = 202) challenge datasets, and an in-house collected dataset (N = 60) were used. Aggregate channel features object detectors were trained to detect reference fat and muscle tissue regions, which were processed and utilized to normalize the 3D images by linear scaling. Mean prostate pseudo T2 values after normalization were compared to literature values. Inter-patient histogram intersections of voxel intensities in the prostate were compared between our approach, the original images, and other commonly used normalization methods. Healthy vs. malignant tissue classification performance was compared before and after normalization. RESULTS: The prostate pseudo T2 values of the three tested datasets (mean ± standard deviation = 78.49 ± 9.42, 79.69 ± 6.34 and 79.29 ± 6.30 ms) corresponded well to T2 values from literature (80 ± 34 ms). Our normalization approach resulted in significantly higher (p < 0.001) inter-patient histogram intersections (median = 0.746) than the original images (median = 0.417) and most other normalization methods. Healthy vs. malignant classification also improved significantly (p < 0.001) in peripheral (AUC 0.826 vs. 0.769) and transition (AUC 0.743 vs. 0.678) zones. CONCLUSION: An automated dual-reference tissue normalization of T2W images could help improve the quantitative assessment of prostate cancer.

Relative Enhanced Diffusivity in Prostate Cancer: Protocol Optimization and Diagnostic Potential.

BACKGROUND: Relative enhanced diffusivity (RED) is a potential biomarker for indirectly measuring perfusion in tissue using diffusion-weighted magnetic resonance imaging (MRI) with 3 b values. PURPOSE: To optimize the RED MRI protocol for the prostate, and to investigate its potential for prostate cancer (PCa) diagnosis. STUDY TYPE: Prospective. POPULATION: Ten asymptomatic healthy volunteers and 35 patients with clinical suspicion of PCa. SEQUENCE: 3T T2 - and diffusion-weighted MRI with b values: b = 0, 50, [100], 150, [200], 250, [300], 400, 800 s/mm2 (values in brackets were only used for patients). ASSESSMENT: Monte Carlo simulations were performed to assess noise sensitivity of RED as a function of intermediate b value. Volunteers were scanned 3 times to assess repeatability of RED. Patient data were used to investigate RED's potential for discriminating between biopsy-confirmed cancer and healthy tissue, and between true and false positive radiological findings. STATISTICAL TESTS: Within-subject coefficient of variation (WCV) to assess repeatability and receiver-operating characteristic curve analysis and logistic regression to assess diagnostic performance of RED. RESULTS: The repeatability was acceptable (WCV = 0.2-0.3) for all intermediate b values tested, apart from b = 50 s/mm2 (WCV = 0.3-0.4). The simulated RED values agreed well with the experimental data, showing that an intermediate b value between 150-250 s/mm2 minimizes noise sensitivity in both peripheral zone (PZ) and transition zone (TZ). RED calculated with the b values 0, 150 and 800 s/mm2 was significantly higher in tumors than in healthy tissue in both PZ (P < 0.001, area under the curve [AUC] = 0.85) and PZ + TZ (P < 0.001, AUC = 0.84). RED was shown to aid apparent diffusion coefficient (ADC) in differentiating between false-positive findings and true-positive PCa in the PZ (AUC; RED = 0.71, ADC = 0.74, RED+ADC = 0.77). DATA CONCLUSION: RED is a repeatable biomarker that may have value for prostate cancer diagnosis. An intermediate b value in the range of 150-250 s/mm2 minimizes the influence of noise and maximizes repeatability. LEVEL OF EVIDENCE: 2 Technical Efficacy Stage: 1 J. Magn. Reson. Imaging 2020;51:1900-1910.

Geometric distortion correction in prostate diffusion-weighted MRI and its effect on quantitative apparent diffusion coefficient analysis.

PURPOSE: To evaluate the effect of correction for B0 inhomogeneity-induced geometric distortion in echo-planar diffusion-weighted imaging on quantitative apparent diffusion coefficient (ADC) analysis in multiparametric prostate MRI. METHODS: Geometric distortion correction was performed in echo-planar diffusion-weighted images (b = 0, 50, 400, 800 s/mm2 ) of 28 patients, using two b0 scans with opposing phase-encoding polarities. Histology-matched tumor and healthy tissue volumes of interest delineated on T2 -weighted images were mapped to the nondistortion-corrected and distortion-corrected data sets by resampling with and without spatial coregistration. The ADC values were calculated on the volume and voxel level. The effect of distortion correction on ADC quantification and tissue classification was evaluated using linear-mixed models and logistic regression, respectively. RESULTS: Without coregistration, the absolute differences in tumor ADC (range: 0.0002-0.189 mm2 /s×10-3 (volume level); 0.014-0.493 mm2 /s×10-3 (voxel level)) between the nondistortion-corrected and distortion-corrected were significantly associated (P < 0.05) with distortion distance (mean: 1.4 ± 1.3 mm; range: 0.3-5.3 mm). No significant associations were found upon coregistration; however, in patients with high rectal gas residue, distortion correction resulted in improved spatial representation and significantly better classification of healthy versus tumor voxels (P < 0.05). CONCLUSIONS: Geometric distortion correction in DWI could improve quantitative ADC analysis in multiparametric prostate MRI. Magn Reson Med 79:2524-2532, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

18F-Fluciclovine PET/MRI for preoperative lymph node staging in high-risk prostate cancer patients.

OBJECTIVE: To investigate the diagnostic potential of simultaneous 18F-fluciclovine PET/MRI for pelvic lymph node (LN) staging in patients with high-risk prostate cancer. METHODS: High-risk prostate cancer patients (n=28) underwent simultaneous 18F-fluciclovine PET/MRI prior to surgery. LNs were removed according to a predefined template of eight regions. PET and MR images were evaluated for presence of LN metastases according to these regions. Sensitivity/specificity for detection of LN metastases were calculated on patient and region basis. Sizes of LN metastases in regions with positive and negative imaging findings were compared with linear mixed models. Clinical parameters of PET-positive and -negative stage N1 patients were compared with the Mann-Whitney U test. RESULTS: Patient- and region-based sensitivity/specificity for detection of pelvic LN metastases was 40 %/87.5 % and 35 %/95.7 %, respectively, for MRI and 40 %/100 % and 30 %/100 %, respectively, for PET. LN metastases in true-positive regions were significantly larger than metastases in false-negative regions. PET-positive stage N1 patients had higher metastatic burden than PET-negative N1 patients. CONCLUSION: Simultaneous 18F-fluciclovine PET/MRI provides high specificity but low sensitivity for detection of LN metastases in high-risk prostate cancer patients. 18F-Fluciclovine PET/MRI scan positive for LN metastases indicates higher metastatic burden than negative scan. KEY POINTS: • 18F-Fluciclovine PET/MRI has high specificity for detection of lymph node metastasis. • 18F-Fluciclovine PET/MRI lacks sensitivity to replace ePLND. • 18F-Fluciclovine PET/MRI may be used to aid surgery and select adjuvant therapy. • 18F-Fluciclovine PET-positive patients have more extensive disease than PET-negative patients. • Size of metastatic lymph nodes is an important factor for detection.

A PET/MRI study towards finding the optimal [18F]Fluciclovine PET protocol for detection and characterisation of primary prostate cancer.

PURPOSE: [18F]Fluciclovine PET imaging shows promise for the assessment of prostate cancer. The purpose of this PET/MRI study is to optimise the PET imaging protocol for detection and characterisation of primary prostate cancer, by quantitative evaluation of the dynamic uptake of [18F]Fluciclovine in cancerous and benign tissue. METHODS: Patients diagnosed with high-risk primary prostate cancer underwent an integrated [18F]Fluciclovine PET/MRI exam before robot-assisted radical prostatectomy with extended pelvic lymph node dissection. Volumes-of-interest (VOIs) of selected organs (prostate, bladder, blood pool) and sub-glandular prostate structures (tumour, benign prostatic hyperplasia (BPH), inflammation, healthy tissue) were delineated on T2-weighted MR images, using whole-mount histology samples as a reference. Three candidate windows for optimal PET imaging were identified based on the dynamic curves of the mean and maximum standardised uptake value (SUVmean and SUVmax, respectively). The statistical significance of differences in SUV between VOIs were analysed using Wilcoxon rank sum tests (p<0.05, adjusted for multiple testing). RESULTS: Twenty-eight (28) patients [median (range) age: 66 (55-72) years] were included. An early (W1: 5-10 minutes post-injection) and two late candidate windows (W2: 18-23; W3: 33-38 minutes post-injection) were selected. Late compared with early imaging was better able to distinguish between malignant and benign tissue [W3, SUVmean: tumour vs. BPH 2.5 vs. 2.0 (p<0.001), tumour vs. inflammation 2.5 vs. 1.7 (p<0.001), tumour vs. healthy tissue 2.5 vs. 2.0 (p<0.001); W1, SUVmean: tumour vs. BPH 3.1 vs. 3.1 (p=0.771), tumour vs inflammation 3.1 vs. 2.2 (p=0.021), tumour vs. healthy tissue 3.1 vs. 2.5 (p<0.001)] as well as between high-grade and low/intermediate-grade tumours (W3, SUVmean: 2.6 vs. 2.1 (p=0.040); W1, SUVmean: 3.1 vs. 2.8 (p=0.173)). These differences were relevant to the peripheral zone, but not the central gland. CONCLUSION: Late-window [18F]Fluciclovine PET imaging shows promise for distinguishing between prostate tumours and benign tissue and for assessment of tumour aggressiveness.

T2-weighted MRI-derived textural features reflect prostate cancer aggressiveness: preliminary results.

PURPOSE: To evaluate the diagnostic relevance of T2-weighted (T2W) MRI-derived textural features relative to quantitative physiological parameters derived from diffusion-weighted (DW) and dynamic contrast-enhanced (DCE) MRI in Gleason score (GS) 3+4 and 4+3 prostate cancers. MATERIALS AND METHODS: 3T multiparametric-MRI was performed on 23 prostate cancer patients prior to prostatectomy. Textural features [angular second moment (ASM), contrast, correlation, entropy], apparent diffusion coefficient (ADC), and DCE pharmacokinetic parameters (Ktrans and Ve) were calculated from index tumours delineated on the T2W, DW, and DCE images, respectively. The association between the textural features and prostatectomy GS and the MRI-derived parameters, and the utility of the parameters in differentiating between GS 3+4 and 4+3 prostate cancers were assessed statistically. RESULTS: ASM and entropy correlated significantly (p < 0.05) with both GS and median ADC. Contrast correlated moderately with median ADC. The textural features correlated insignificantly with Ktrans and Ve. GS 4+3 cancers had significantly lower ASM and higher entropy than 3+4 cancers, but insignificant differences in median ADC, Ktrans, and Ve. The combined texture-MRI parameters yielded higher classification accuracy (91%) than the individual parameter sets. CONCLUSION: T2W MRI-derived textural features could serve as potential diagnostic markers, sensitive to the pathological differences in prostate cancers. KEY POINTS: • T2W MRI-derived textural features correlate significantly with Gleason score and ADC. • T2W MRI-derived textural features differentiate Gleason score 3+4 from 4+3 cancers. • T2W image textural features could augment tumour characterization.

Toward Simultaneous Real-Time Fluoroscopic and Nuclear Imaging in the Intervention Room.

PURPOSE: To investigate the technical feasibility of hybrid simultaneous fluoroscopic and nuclear imaging. MATERIALS AND METHODS: An x-ray tube, an x-ray detector, and a gamma camera were positioned in one line, enabling imaging of the same field of view. Since a straightforward combination of these elements would block the lines of view, a gamma camera setup was developed to be able to view around the x-ray tube. A prototype was built by using a mobile C-arm and a gamma camera with a four-pinhole collimator. By using the prototype, test images were acquired and sensitivity, resolution, and coregistration error were analyzed. RESULTS: Nuclear images (two frames per second) were acquired simultaneously with fluoroscopic images. Depending on the distance from point source to detector, the system resolution was 1.5-1.9-cm full width at half maximum, the sensitivity was (0.6-1.5) × 10(-5) counts per decay, and the coregistration error was -0.13 to 0.15 cm. With good spatial and temporal alignment of both modalities throughout the field of view, fluoroscopic images can be shown in grayscale and corresponding nuclear images in color overlay. CONCLUSION: Measurements obtained with the hybrid imaging prototype device that combines simultaneous fluoroscopic and nuclear imaging of the same field of view have demonstrated the feasibility of real-time simultaneous hybrid imaging in the intervention room.

(99m)Tc-MAA overestimates the absorbed dose to the lungs in radioembolization: a quantitative evaluation in patients treated with ¹66Ho-microspheres.

PURPOSE: Radiation pneumonitis is a rare but serious complication of radioembolic therapy of liver tumours. Estimation of the mean absorbed dose to the lungs based on pretreatment diagnostic (99m)Tc-macroaggregated albumin ((99m)Tc-MAA) imaging should prevent this, with administered activities adjusted accordingly. The accuracy of (99m)Tc-MAA-based lung absorbed dose estimates was evaluated and compared to absorbed dose estimates based on pretreatment diagnostic (166)Ho-microsphere imaging and to the actual lung absorbed doses after (166)Ho radioembolization. METHODS: This prospective clinical study included 14 patients with chemorefractory, unresectable liver metastases treated with (166)Ho radioembolization. (99m)Tc-MAA-based and (166)Ho-microsphere-based estimation of lung absorbed doses was performed on pretreatment diagnostic planar scintigraphic and SPECT/CT images. The clinical analysis was preceded by an anthropomorphic torso phantom study with simulated lung shunt fractions of 0 to 30 % to determine the accuracy of the image-based lung absorbed dose estimates after (166)Ho radioembolization. RESULTS: In the phantom study, (166)Ho SPECT/CT-based lung absorbed dose estimates were more accurate (absolute error range 0.1 to -4.4 Gy) than (166)Ho planar scintigraphy-based lung absorbed dose estimates (absolute error range 9.5 to 12.1 Gy). Clinically, the actual median lung absorbed dose was 0.02 Gy (range 0.0 to 0.7 Gy) based on posttreatment (166)Ho-microsphere SPECT/CT imaging. Lung absorbed doses estimated on the basis of pretreatment diagnostic (166)Ho-microsphere SPECT/CT imaging (median 0.02 Gy, range 0.0 to 0.4 Gy) were significantly better predictors of the actual lung absorbed doses than doses estimated on the basis of (166)Ho-microsphere planar scintigraphy (median 10.4 Gy, range 4.0 to 17.3 Gy; p < 0.001), (99m)Tc-MAA SPECT/CT imaging (median 2.5 Gy, range 1.2 to 12.3 Gy; p < 0.001), and (99m)Tc-MAA planar scintigraphy (median 5.5 Gy, range 2.3 to 18.2 Gy; p < 0.001). CONCLUSION: In clinical practice, lung absorbed doses are significantly overestimated by pretreatment diagnostic (99m)Tc-MAA imaging. Pretreatment diagnostic (166)Ho-microsphere SPECT/CT imaging accurately predicts lung absorbed doses after (166)Ho radioembolization.

Radiation-induced cholecystitis after hepatic radioembolization: do we need to take precautionary measures?

Controversy exists over the need to take precautionary measures during hepatic radioembolization to minimize the risk of radiation-induced cholecystitis. Strategies for a variety of clinical scenarios are discussed on the basis of a literature review. Precautionary measures are unnecessary in the majority of patients and should be taken only when single photon-emission computed tomography (CT; SPECT)/CT shows a significant concentration of technetium-99m macroaggregated albumin in the gallbladder wall. In this case report with quantitative SPECT analysis, it is illustrated how an adjustment of the catheter position can effectively reduce the absorbed dose of radiation delivered to the gallbladder wall by more than 90%.

Repeat Prostate-specific Antigen Testing Improves Risk-based Selection of Men for Prostate Biopsy After Magnetic Resonance Imaging.

Background and objective: The aim of our study was to investigate whether repeat prostate-specific antigen (PSA) testing as currently recommended improves risk stratification for men undergoing magnetic resonance imaging (MRI) and targeted biopsy for suspected prostate cancer (PCa). Methods: Consecutive men undergoing MRI and prostate biopsy who had at least two PSA tests before prostate biopsy were retrospectively registered and assigned to a development cohort (n = 427) or a validation (n = 174) cohort. Change in PSA level was assessed as a predictor of clinically significant PCa (csPCa; Gleason score ≥3 + 4, grade group ≥2) by multivariable logistic regression analysis. We developed a multivariable prediction model (MRI-RC) and a dichotomous biopsy decision strategy incorporating the PSA change. The performance of the MRI-RC model and dichotomous decision strategy was assessed in the validation cohort and compared to prediction models and decision strategies not including PSA change in terms of discriminative ability and decision curve analysis. Results: Men who had a decrease on repeat PSA testing had significantly lower risk of csPCa than men without a decrease (odds ratio [OR] 0.3, 95% confidence interval [CI] 0.16-0.54; p < 0.001). Men with an increased repeat PSA had a significantly higher risk of csPCa than men without an increase (OR 2.97, 95% CI 1.62-5.45; p < 0.001). Risk stratification using both the MRI-RC model and the dichotomous decision strategy was improved by incorporating change in PSA as a parameter. Conclusions and clinical implications: Repeat PSA testing gives predictive information regarding men undergoing MRI and targeted prostate biopsy. Inclusion of PSA change as a parameter in an MRI-RC model and a dichotomous biopsy decision strategy improves their predictive performance and clinical utility without requiring additional investigations. Patient summary: For men with a suspicion of prostate cancer, repeat PSA (prostate-specific antigen) testing after an MRI (magnetic resonance imaging) scan can help in identifying patients who can safely avoid prostate biopsy.

MRI-based biodistribution assessment of holmium-166 poly(L-lactic acid) microspheres after radioembolisation.

OBJECTIVES: To demonstrate the feasibility of MRI-based assessment of the intrahepatic Ho-PLLA-MS biodistribution after radioembolisation in order to estimate the absorbed radiation dose. METHODS: Fifteen patients were treated with holmium-166 ((166)Ho) poly(L-lactic acid)-loaded microspheres (Ho-PLLA-MS, mean 484 mg; range 408-593 mg) in a phase I study. Multi-echo gradient-echo MR images were acquired from which R (2) maps were constructed. The amount of Ho-PLLA-MS in the liver was determined by using the relaxivity r (2) of the Ho-PLLA-MS and compared with the administered amount. Quantitative single photon emission computed tomography (SPECT) was used for comparison with MRI regarding the whole liver absorbed radiation dose. RESULTS: R (2) maps visualised the deposition of Ho-PLLA-MS with great detail. The mean total amount of Ho-PLLA-MS detected in the liver based on MRI was 431 mg (range 236-666 mg) or 89 ± 19 % of the delivered amount (correlation coefficient r = 0.7; P < 0.01). A good correlation was found between the whole liver mean absorbed radiation dose as assessed by MRI and SPECT (correlation coefficient r = 0.927; P < 0.001). CONCLUSION: MRI-based dosimetry for holmium-166 radioembolisation is feasible. Biodistribution is visualised with great detail and quantitative measurements are possible.

CROPro: a tool for automated cropping of prostate magnetic resonance images.

Purpose: To bypass manual data preprocessing and optimize deep learning performance, we developed and evaluated CROPro, a tool to standardize automated cropping of prostate magnetic resonance (MR) images. Approach: CROPro enables automatic cropping of MR images regardless of patient health status, image size, prostate volume, or pixel spacing. CROPro can crop foreground pixels from a region of interest (e.g., prostate) with different image sizes, pixel spacing, and sampling strategies. Performance was evaluated in the context of clinically significant prostate cancer (csPCa) classification. Transfer learning was used to train five convolutional neural network (CNN) and five vision transformer (ViT) models using different combinations of cropped image sizes ( 64 × 64 , 128 × 128 , and 256 × 256  pixels2), pixel spacing ( 0.2 × 0.2 , 0.3 × 0.3 , 0.4 × 0.4 , and 0.5 × 0.5 mm 2 ), and sampling strategies (center, random, and stride cropping) over the prostate. T2-weighted MR images ( N = 1475 ) from the online available PI-CAI challenge were used to train ( N = 1033 ), validate ( N = 221 ), and test ( N = 221 ) all models. Results: Among CNNs, SqueezeNet with stride cropping (image size: 128 × 128 , pixel spacing: 0.2 × 0.2 mm 2 ) achieved the best classification performance ( 0.678 ± 0.006 ). Among ViTs, ViT-H/14 with random cropping (image size: 64 × 64 and pixel spacing: 0.5 × 0.5 mm 2 ) achieved the best performance ( 0.756 ± 0.009 ). Model performance depended on the cropped area, with optimal size generally larger with center cropping ( ∼ 40 cm 2 ) than random/stride cropping ( ∼ 10 cm 2 ). Conclusion: We found that csPCa classification performance of CNNs and ViTs depends on the cropping settings. We demonstrated that CROPro is well suited to optimize these settings in a standardized manner, which could improve the overall performance of deep learning models.

Label-set impact on deep learning-based prostate segmentation on MRI.

BACKGROUND: Prostate segmentation is an essential step in computer-aided detection and diagnosis systems for prostate cancer. Deep learning (DL)-based methods provide good performance for prostate gland and zones segmentation, but little is known about the impact of manual segmentation (that is, label) selection on their performance. In this work, we investigated these effects by obtaining two different expert label-sets for the PROSTATEx I challenge training dataset (n = 198) and using them, in addition to an in-house dataset (n = 233), to assess the effect on segmentation performance. The automatic segmentation method we used was nnU-Net. RESULTS: The selection of training/testing label-set had a significant (p < 0.001) impact on model performance. Furthermore, it was found that model performance was significantly (p < 0.001) higher when the model was trained and tested with the same label-set. Moreover, the results showed that agreement between automatic segmentations was significantly (p < 0.0001) higher than agreement between manual segmentations and that the models were able to outperform the human label-sets used to train them. CONCLUSIONS: We investigated the impact of label-set selection on the performance of a DL-based prostate segmentation model. We found that the use of different sets of manual prostate gland and zone segmentations has a measurable impact on model performance. Nevertheless, DL-based segmentation appeared to have a greater inter-reader agreement than manual segmentation. More thought should be given to the label-set, with a focus on multicenter manual segmentation and agreement on common procedures. CRITICAL RELEVANCE STATEMENT: Label-set selection significantly impacts the performance of a deep learning-based prostate segmentation model. Models using different label-set showed higher agreement than manual segmentations. KEY POINTS: • Label-set selection has a significant impact on the performance of automatic segmentation models. • Deep learning-based models demonstrated true learning rather than simply mimicking the label-set. • Automatic segmentation appears to have a greater inter-reader agreement than manual segmentation.