Search

Objective The purpose of this study was: To test whether machine learning classifiers for transition zone (TZ) and peripheral zone (PZ) can correctly classify prostate tumors into those with/without a Gleason 4 component, and to compare the performance of the best performing classifiers against the opinion of three board-certified radiologists. Methods A retrospective analysis of prospectively acquired data was performed at a single center between 2012 and 2015. Inclusion criteria were (i) 3-T mp-MRI compliant with international guidelines, (ii) Likert ≥ 3/5 lesion, (iii) transperineal template ± targeted index lesion biopsy confirming cancer ≥ Gleason 3 + 3. Index lesions from 164 men were analyzed (119 PZ, 45 TZ). Quantitative MRI and clinical features were used and zone-specific machine learning classifiers were constructed. Models were validated using a fivefold cross-validation and a temporally separated patient cohort. Classifier performance was compared against the opinion of three board-certified radiologists. Results The best PZ classifier trained with prostate-specific antigen density, apparent diffusion coefficient (ADC), and maximum enhancement (ME) on DCE-MRI obtained a ROC area under the curve (AUC) of 0.83 following fivefold cross-validation. Diagnostic sensitivity at 50% threshold of specificity was higher for the best PZ model (0.93) when compared with the mean sensitivity of the three radiologists (0.72). The best TZ model used ADC and ME to obtain an AUC of 0.75 following fivefold cross-validation. This achieved higher diagnostic sensitivity at 50% threshold of specificity (0.88) than the mean sensitivity of the three radiologists (0.82). Conclusions Machine learning classifiers predict Gleason pattern 4 in prostate tumors better than radiologists. Key Points • Predictive models developed from quantitative multiparametric magnetic resonance imaging regarding the characterization of prostate cancer grade should be zone-specific. • Classifiers trained differently for peripheral and transition zone can predict a Gleason 4 component with a higher performance than the subjective opinion of experienced radiologists. • Classifiers would be particularly useful in the context of active surveillance, whereby decisions regarding whether to biopsy are necessitated.

Background Multiparametric magnetic resonance imaging (mp-MRI) is becoming an increasingly important diagnostic tool for prostate cancer. So far there has been little focus on management for indeterminate mp-MRI results. Objective To describe outcomes for a cohort of men rated as having an indeterminate mp-MRI result. Design, setting, and participants Patients were identified retrospectively from a single UK centre between October 2010 and January 2015. Patients were included if they had a Likert score of 3/5 on a first MRI scan without any prior prostate biopsy. Patients were offered one of two initial management strategies. Strategy 1 was an immediate targeted biopsy of the MRI lesion. Strategy 2 was a surveillance process comprising prostate-specific antigen monitoring and/or mp-MRI at intervals of 6–12 mo, with biopsy on a for-cause basis. Outcome measurements and statistical analysis Cancer detection and treatment outcomes were compared for the two strategies. Results and limitations Of 168 patients, 73 (43%) chose strategy 1 and 95 (57%) chose strategy two. The overall proportion of men with clinically significant cancer detected was 14% (23/168). The risk profile for cancer identified in the initial surveillance group was similar to that identified in the immediate biopsy group. Limitations of the study include the short follow-up. Conclusions Men with indeterminate mp-MRI were willing to forego immediate biopsy for a strategy of surveillance involving PSA measurement and/or mp-MRI repeated at intervals. The risk profile of the cancers identified by both strategies appeared similar, but many men in the surveillance group avoided the risks, complications, and costs of biopsy. Long-term results are awaited. Patient summary This report compares two approaches for an uncertain magnetic resonance imaging result for clinically important prostate cancer: immediate biopsy versus surveillance with delayed biopsy if required. Delayed biopsy did not result in identification of cancer with adverse features, and many men benefited from avoiding a biopsy and its complications.

Objectives: To assess the clinical outcomes of mpMRI before biopsy and evaluate the space remaining for novel biomarkers. Methods: The INNOVATE study was set up to evaluate the validity of novel fluidic biomarkers in men with suspected prostate cancer who undergo pre-biopsy mpMRI. We report the characteristics of this clinical cohort, the distribution of clinical serum biomarkers, PSA and PSA density (PSAD), and compare the mpMRI Likert scoring system to the Prostate Imaging–Reporting and Data System v2.1 (PI-RADS) in men undergoing biopsy. Results: 340 men underwent mpMRI to evaluate suspected prostate cancer. 193/340 (57%) men had subsequent MRI-targeted prostate biopsy. Clinically significant prostate cancer (csigPCa), i.e., overall Gleason ≥ 3 + 4 of any length OR maximum cancer core length (MCCL) ≥4 mm of any grade including any 3 + 3, was found in 96/195 (49%) of biopsied patients. Median PSA (and PSAD) was 4.7 (0.20), 8.0 (0.17), and 9.7 (0.31) ng/mL (ng/mL/mL) in mpMRI scored Likert 3,4,5 respectively for men with csigPCa on biopsy. The space for novel biomarkers was shown to be within the group of men with mpMRI scored Likert3 (178/340) and 4 (70/350), in whom an additional of 40% (70/178) men with mpMRI-scored Likert3, and 37% (26/70) Likert4 could have been spared biopsy. PSAD is already considered clinically in this cohort to risk stratify patients for biopsy, despite this 67% (55/82) of men with mpMRI-scored Likert3, and 55% (36/65) Likert4, who underwent prostate biopsy had a PSAD below a clinical threshold of 0.15 (or 0.12 for men aged &lt, 50 years). Different thresholds of PSA and PSAD were assessed in mpMRI-scored Likert4 to predict csigPCa on biopsy, to achieve false negative levels of ≤5% the proportion of patients whom who test as above the threshold were unsuitably high at 86 and 92% of patients for PSAD and PSA respectively. When PSA was re tested in a sub cohort of men repeated PSAD showed its poor reproducibility with 43% (41/95) of patients being reclassified. After PI-RADS rescoring of the biopsied lesions, 66% (54/82) of the Likert3 lesions received a different PI-RADS score. Conclusions: The addition of simple biochemical and radiological markers (Likert and PSAD) facilitate the streamlining of the mpMRI-diagnostic pathway for suspected prostate cancer but there remains scope for improvement, in the introduction of novel biomarkers for risk assessment in Likert3 and 4 patients, future application of novel biomarkers tested in a Likert cohort would also require re-optimization around Likert3/PI-RADS2, as well as reproducibility testing.

The UK National Institute for Health and Care Excellence (NICE) 2019 "Prostate cancer: diagnosis and management" guidelines have recommended that all patients suspected of prostate cancer undergo multiparametric magnetic resonance imaging (mpMRI) prior to biopsy. The Likert scoring system is advocated for mpMRI reporting based on multicentre studies that have demonstrated its effectiveness within the National Health Service (NHS). In recent years, there has been considerable drive towards standardised prostate reporting, which led to the development of "Prostate Imaging-Reporting And Data System" (PI-RADS). The PI-RADS system has been adopted by the majority of European countries and within the US. This paper reviews these systems indicating the similarities and specific differences that exist between PI-RADS and Likert assessment through a series of histologically proven clinical cases.

The VERDICT framework for modelling diffusion MRI data aims to relate parameters from a biophysical model to histological features used for tumour grading in prostate cancer. Validation of the VERDICT model is necessary for clinical use. This study compared VERDICT parameters obtained ex vivo with histology in five specimens from radical prostatectomy. A patient-specific 3D-printed mould was used to investigate the effects of fixation on VERDICT parameters and to aid registration to histology. A rich diffusion data set was acquired in each ex vivo prostate before and after fixation. At both time points, data were best described by a two-compartment model: the model assumes that an anisotropic tensor compartment represents the extracellular space and a restricted sphere compartment models the intracellular space. The effect of fixation on model parameters associated with tissue microstructure was small. The patient-specific mould minimized tissue deformations and co-localized slices, so that rigid registration of MRI to histology images allowed region-based comparison with histology. The VERDICT estimate of the intracellular volume fraction corresponded to histological indicators of cellular fraction, including high values in tumour regions. The average sphere radius from VERDICT, representing the average cell size, was relatively uniform across samples. The primary diffusion direction from the extracellular compartment of the VERDICT model aligned with collagen fibre patterns in the stroma obtained by structure tensor analysis. This confirmed the biophysical relationship between ex vivo VERDICT parameters and tissue microstructure from histology.

Background Biologic specificity of diffusion MRI in relation to prostate cancer aggressiveness may improve by examining separate components of the diffusion MRI signal. The Vascular, Extracellular, and Restricted Diffusion for Cytometry in Tumors (VERDICT) model estimates three distinct signal components and associates them to (a) intracellular water, (b) water in the extracellular extravascular space, and (c) water in the microvasculature. Purpose To evaluate the repeatability, image quality, and diagnostic utility of intracellular volume fraction (FIC) maps obtained with VERDICT prostate MRI and to compare those maps with apparent diffusion coefficient (ADC) maps for Gleason grade differentiation. Materials and Methods Seventy men (median age, 62.2 years; range, 49.5-82.0 years) suspected of having prostate cancer or undergoing active surveillance were recruited to a prospective study between April 2016 and October 2017. All men underwent multiparametric prostate and VERDICT MRI. Forty-two of the 70 men (median age, 67.7 years; range, 50.0-82.0 years) underwent two VERDICT MRI acquisitions to assess repeatability of FIC measurements obtained with VERDICT MRI. Repeatability was measured with use of intraclass correlation coefficients (ICCs). The image quality of FIC and ADC maps was independently evaluated by two board-certified radiologists. Forty-two men (median age, 64.8 years; range, 49.5-79.6 years) underwent targeted biopsy, which enabled comparison of FIC and ADC metrics in the differentiation between Gleason grades. Results VERDICT MRI FIC demonstrated ICCs of 0.87-0.95. There was no significant difference between image quality of ADC and FIC maps (score, 3.1 vs 3.3, respectively; P = .90). FIC was higher in lesions with a Gleason grade of at least 3+4 compared with benign and/or Gleason grade 3+3 lesions (mean, 0.49 ± 0.17 vs 0.31 ± 0.12, respectively; P = .002). The difference in ADC between these groups did not reach statistical significance (mean, 1.42 vs 1.16 × 10-3 mm2/sec; P = .26). Conclusion Fractional intracellular volume demonstrates high repeatability and image quality and enables better differentiation of a Gleason 4 component cancer from benign and/or Gleason 3+3 histology than apparent diffusion coefficient. Online supplemental material is available for this article. See also the editorial by Sigmund and Rosenkrantz in this issue.

OBJECTIVES: Compare the performance of zone-specific multi-parametric-MRI (mp-MRI) diagnostic models in prostate cancer detection with experienced radiologists. METHODS: A single-centre, IRB approved, prospective STARD compliant 3 T MRI test dataset of 203 patients was generated to test validity and generalisability of previously reported 1.5 T mp-MRI diagnostic models. All patients included within the test dataset underwent 3 T mp-MRI, comprising T2, diffusion-weighted and dynamic contrast-enhanced imaging followed by transperineal template ± targeted index lesion biopsy. Separate diagnostic models (transition zone (TZ) and peripheral zone (PZ)) were applied to respective zones. Sensitivity/specificity and the area under the receiver operating characteristic curve (ROC-AUC) were calculated for the two zone-specific models. Two radiologists (A and B) independently Likert scored test 3 T mp-MRI dataset, allowing ROC analysis for each radiologist for each prostate zone. RESULTS: Diagnostic models applied to the test dataset demonstrated a ROC-AUC = 0.74 (95% CI 0.67-0.81) in the PZ and 0.68 (95% CI 0.61-0.75) in the TZ. Radiologist A/B had a ROC-AUC = 0.78/0.74 in the PZ and 0.69/0.69 in the TZ. Radiologists A and B each scored 51 patients in the PZ and 41 and 45 patients respectively in the TZ as Likert 3. The PZ model demonstrated a ROC-AUC = 0.65/0.67 for the patients Likert scored as indeterminate by radiologist A/B respectively, whereas the TZ model demonstrated a ROC-AUC = 0.74/0.69. CONCLUSION: Zone-specific mp-MRI diagnostic models demonstrate generalisability between 1.5 and 3 T mp-MRI protocols and show similar classification performance to experienced radiologists for prostate cancer detection. Results also indicate the ability of diagnostic models to classify cases with an indeterminate radiologist score. KEY POINTS: • MRI diagnostic models had similar performance to experienced radiologists for classification of prostate cancer. • MRI diagnostic models may help radiologists classify tumour in patients with indeterminate Likert 3 scores.

To review the use of multi-parametric MRI (mpMRI) in loco-regional assessment of men with early prostate cancer.mpMRI comprises anatomic T2 and T1 sequences supplemented by functional imaging techniques such as diffusion-weighted and dynamic contrast enhanced (DCE) imaging. mpMRI is gaining increasing acceptance for prostate cancer detection and staging of early disease. It can facilitate targeted therapies, guide surgical options and enable active surveillance within suitable patients. The technique can be performed at 1.5 or 3 Tesla, but sequence optimization is critical to successful implementation of mpMRI. T2 and diffusion-weighted sequences are minimal requirements and are often complemented by DCE images. When performed at high spatial resolution, DCE facilitates detection of disease, as well as assessment of extra-capsular extension, distal urethral sphincter and seminal vesicles involvement. Pre-biopsy mpMRI is recommended for both detection and staging as it avoids biopsy artefact, and when normal, has a negative predictive value of 95% for significant cancer.mpMRI reliably detects clinically significant prostate tumour and ideally should be performed prior to biopsy. It provides an accurate method for local disease staging and facilitates a growing range of treatment options for patients with early disease.

Multi-Atlas based Segmentation (MAS) algorithms have been successfully applied to many medical image segmentation tasks, but their success relies on a large number of atlases and good image registration performance. Choosing well-registered atlases for label fusion is vital for an accurate segmentation. This choice becomes even more crucial when the segmentation involves organs characterized by a high anatomical and pathological variability. In this paper, we propose a new genetic atlas selection strategy (GAS) that automatically chooses the best subset of atlases to be used for segmenting the target image, on the basis of both image similarity and segmentation overlap. More precisely, the key idea of GAS is that if two images are similar, the performances of an atlas for segmenting each image are similar. Since the ground truth of each atlas is known, GAS first selects a predefined number of similar images to the target, then, for each one of them, finds a near-optimal subset of atlases by means of a genetic algorithm. All these near-optimal subsets are then combined and used to segment the target image. GAS was tested on single-label and multi-label segmentation problems. In the first case, we considered the segmentation of both the whole prostate and of the left ventricle of the heart from magnetic resonance images. Regarding multi-label problems, the zonal segmentation of the prostate into peripheral and transition zone was considered. The results showed that the performance of MAS algorithms statistically improved when GAS is used.

OBJECTIVE: To determine whether indeterminate (Likert-score 3/5) peripheral zone (PZ) multiparametric MRI (mpMRI) studies are classifiable by prostate-specific antigen (PSA), PSA density (PSAD), Prostate Imaging Reporting And Data System version 2 (PI-RADS_v2) rescoring and morphological MRI features. METHODS: Men with maximum Likert-score 3/5 within their PZ were retrospectively selected from 330 patients who prospectively underwent prostate mpMRI (3 T) without an endorectal coil, followed by 20-zone transperineal template prostate mapping biopsies ＋/- focal lesion-targeted biopsy. PSAD was calculated using pre-biopsy PSA and MRI-derived volume. Two readers A and B independently assessed included men with both Likert-assessment and PI-RADS_v2. Both readers then classified mpMRI morphological features in consensus. Men were divided into two groups: significant cancer (≥ Gleason 3 ＋ 4) or insignificant cancer (≤ Gleason 3 ＋ 3)/no cancer. Comparisons between groups were made separately for PSA & PSAD using Mann–Whitney test and morphological descriptors with Fisher’s exact test. PI-RADS_v2 and Likert-assessment were descriptively compared and percentage inter-reader agreement calculated. RESULTS: 76 males were eligible for PSA & PSAD analyses, 71 for PI-RADS scoring, and 67 for morphological assessment (excluding significant image artefacts). Unlike PSA (p = 0.915), PSAD was statistically different (p = 0.004) between the significant [median: 0.19 ng ml(–)(2) (interquartile range: 0.13–0.29)] and non-significant/no cancer [median: 0.13 ng ml(–)(2) (interquartile range: 0.10–0.17)] groups. Presence of mpMRI morphological features was not significantly different between groups. Subjective Likert-assessment discriminated patients with significant cancer better than PI-RADS_v2. Inter-reader percentage agreement was 83% for subjective Likert-assessment and 56% for PI-RADS_v2. CONCLUSION: PSAD may categorize presence of significant cancer in patients with Likert-scored 3/5 PZ mpMRI findings. ADVANCES IN KNOWLEDGE: PSAD may be used in indeterminate PZ mpMRI to guide decisions between biopsy vs monitoring.

Feuillets de Radiologie - In Press.Proof corrected by the author Available online since mercredi 26 juin 2013