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.

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.

Objectives: To evaluate whole-body MRI (WB-MRI) parameters significantly associated with treatment response in multiple myeloma (MM)., Methods: Twenty-one MM patients underwent WB-MRI at diagnosis and after two cycles of chemotherapy. Scans acquired at 3.0 T included T2, diffusion-weighted-imaging (DWI) and mDixon pre- and post-contrast. Twenty focal lesions (FLs) matched on DWI and post-contrast mDixon were selected for each time point. Estimated tumour volume (eTV), apparent diffusion coefficient (ADC), enhancement ratio (ER) and signal fat fraction (sFF) were derived. Clinical treatment response to chemotherapy was assessed using conventional criteria. Significance of temporal parameter change was assessed by the paired t test and receiver operating characteristics/area under the curve (AUC) analysis was performed. Parameter repeatability was assessed by interclass correlation (ICC) and Bland-Altman analysis of 10 healthy volunteers scanned at two time points., Results: Fifteen of 21 patients responded to treatment. Of 254 FLs analysed, sFF (p < 0.0001) and ADC (p = 0.001) significantly increased in responders but not non-responders. eTV significantly decreased in 19/21 cases. Focal lesion sFF was the best discriminator of treatment response (AUC 1.0). Bone sFF repeatability was excellent (ICC 0.98) and better than bone ADC (ICC 0.47)., Conclusion: WB-MRI derived focal lesion sFF shows promise as an imaging biomarker of treatment response in newly diagnosed MM., Key Points: • Bone signal fat fraction using mDixon is a robust quantifiable parameter • Fat fraction and ADC significantly increase in myeloma lesions responding to treatment • Bone lesion fat fraction is the best discriminator of myeloma treatment response.

Objectives: To evaluate multiparametric-MRI (mpMRI) derived histogram textural-analysis parameters for detection of transition zone (TZ) prostatic tumour., Methods: Sixty-seven consecutive men with suspected prostate cancer underwent 1.5T mpMRI prior to template-mapping-biopsy (TPM). Twenty-six men had 'significant' TZ tumour. Two radiologists in consensus matched TPM to the single axial slice best depicting tumour, or largest TZ diameter for those with benign histology, to define single-slice whole TZ-regions-of-interest (ROIs). Textural-parameter differences between single-slice whole TZ-ROI containing significant tumour versus benign/insignificant tumour were analysed using Mann Whitney U test. Diagnostic accuracy was assessed by receiver operating characteristic area under curve (ROC-AUC) analysis cross-validated with leave-one-out (LOO) analysis., Results: ADC kurtosis was significantly lower (p < 0.001) in TZ containing significant tumour with ROC-AUC 0.80 (LOO-AUC 0.78); the difference became non-significant following exclusion of significant tumour from single-slice whole TZ-ROI (p = 0.23). T1-entropy was significantly lower (p = 0.004) in TZ containing significant tumour with ROC-AUC 0.70 (LOO-AUC 0.66) and was unaffected by excluding significant tumour from TZ-ROI (p = 0.004). Combining these parameters yielded ROC-AUC 0.86 (LOO-AUC 0.83)., Conclusion: Textural features of the whole prostate TZ can discriminate significant prostatic cancer through reduced kurtosis of the ADC-histogram where significant tumour is included in TZ-ROI and reduced T1 entropy independent of tumour inclusion., Key Points: • MR textural features of prostate transition zone may discriminate significant prostatic cancer. • Transition zone (TZ) containing significant tumour demonstrates a less peaked ADC histogram. • TZ containing significant tumour reveals higher post-contrast T1-weighted homogeneity. • The utility of MR texture analysis in prostate cancer merits further investigation.

Objectives: To assess the interchangeability of zone-specific (peripheral-zone (PZ) and transition-zone (TZ)) multiparametric-MRI (mp-MRI) logistic-regression (LR) models for classification of prostate cancer., Methods: Two hundred and thirty-one patients (70 TZ training-cohort; 76 PZ training-cohort; 85 TZ temporal validation-cohort) underwent mp-MRI and transperineal-template-prostate-mapping biopsy. PZ and TZ uni/multi-variate mp-MRI LR-models for classification of significant cancer (any cancer-core-length (CCL) with Gleason > 3 + 3 or any grade with CCL ≥ 4 mm) were derived from the respective cohorts and validated within the same zone by leave-one-out analysis. Inter-zonal performance was tested by applying TZ models to the PZ training-cohort and vice-versa. Classification performance of TZ models for TZ cancer was further assessed in the TZ validation-cohort. ROC area-under-curve (ROC-AUC) analysis was used to compare models., Results: The univariate parameters with the best classification performance were the normalised T2 signal (T2nSI) within the TZ (ROC-AUC = 0.77) and normalized early contrast-enhanced T1 signal (DCE-nSI) within the PZ (ROC-AUC = 0.79). Performance was not significantly improved by bi-variate/tri-variate modelling. PZ models that contained DCE-nSI performed poorly in classification of TZ cancer. The TZ model based solely on maximum-enhancement poorly classified PZ cancer., Conclusion: LR-models dependent on DCE-MRI parameters alone are not interchangable between prostatic zones; however, models based exclusively on T2 and/or ADC are more robust for inter-zonal application., Key Points: • The ADC and T2-nSI of benign/cancer PZ are higher than benign/cancer TZ. • DCE parameters are significantly different between benign PZ and TZ, but not between cancerous PZ and TZ. • Diagnostic models containing contrast enhancement parameters have reduced performance when applied across zones.

Objectives: We aimed to develop logistic regression (LR) models for classifying prostate cancer within the transition zone on multi-parametric magnetic resonance imaging (mp-MRI)., Methods: One hundred and fifty-five patients (training cohort, 70 patients; temporal validation cohort, 85 patients) underwent mp-MRI and transperineal-template-prostate-mapping (TPM) biopsy. Positive cores were classified by cancer definitions: (1) any-cancer; (2) definition-1 [≥Gleason 4 + 3 or ≥ 6 mm cancer core length (CCL)] [high risk significant]; and (3) definition-2 (≥Gleason 3 + 4 or ≥ 4 mm CCL) cancer [intermediate-high risk significant]. For each, logistic-regression mp-MRI models were derived from the training cohort and validated internally and with the temporal cohort. Sensitivity/specificity and the area under the receiver operating characteristic (ROC-AUC) curve were calculated. LR model performance was compared to radiologists' performance., Results: Twenty-eight of 70 patients from the training cohort, and 25/85 patients from the temporal validation cohort had significant cancer on TPM. The ROC-AUC of the LR model for classification of cancer was 0.73/0.67 at internal/temporal validation. The radiologist A/B ROC-AUC was 0.65/0.74 (temporal cohort). For patients scored by radiologists as Prostate Imaging Reporting and Data System (Pi-RADS) score 3, sensitivity/specificity of radiologist A 'best guess' and LR model was 0.14/0.54 and 0.71/0.61, respectively; and radiologist B 'best guess' and LR model was 0.40/0.34 and 0.50/0.76, respectively., Conclusions: LR models can improve classification of Pi-RADS score 3 lesions similar to experienced radiologists., Key Points: • MRI helps find prostate cancer in the anterior of the gland • Logistic regression models based on mp-MRI can classify prostate cancer • Computers can help confirm cancer in areas doctors are uncertain about.

Objectives: To develop an MRI enterography global score (MEGS) of Crohn's disease (CD) activity compared with a reference standard of faecal calprotectin (fC), C-reactive protein (CRP) and Harvey-Bradshaw index (HBI)., Methods: Calprotectin, CRP and HBI were prospectively recorded for 71 patients (median age 33, male 35) with known/suspected CD undergoing MRI enterography. Two observers in consensus scored activity for nine bowel segments, grading mural thickness, T2 signal, mesenteric oedema, T1 enhancement and pattern, and haustral loss. Segmental scores were multiplied according to disease length. Five points each were added for lymphadenopathy, comb sign, fistulae and abscesses to derive the MEGS. A previously validated MRI CD activity score (CDAS) was also calculated. MRI scores were correlated with clinical references using Spearman's rank. A logistic regression diagnostic model was built to discriminate active (fC > 100 μg/g) from inactive disease., Results: MEGS and CDAS were significantly correlated with fC (r = 0.46, P < 0.001) and (r = 0.39, P = 0.001) respectively. MEGS correlated with CRP (r = 0.39, P = 0.002). The model for discriminating active from inactive disease achieved an area under the receiver-operating curve of 0.75 and 0.66 after leave-one-out analysis., Conclusion: A magnetic resonance enterography global score (MEGS) of CD activity correlated significantly with fC levels., Key Points: • Magnetic resonance imaging is now widely used to assess Crohn's disease. • Existing MRI activity scores depend on local segmental endoscopic/histological reference standards. • Scores including assessment of disease extent/complications better demonstrate full disease burden. • This new global Crohn's disease burden score correlates with calprotectin and CRP. • The MRI enterography score of disease activity can complement existing clinical markers.

Objectives: Magnetic resonance imaging (MRI) acquired on equipment capable of simultaneous MRI and positron emission tomography (PET) could potentially provide the gold standard method for motion correction of PET. To assess the latter, in this study we compared fast 2D and 3D MRI of the torso and used deformation parameters from real MRI data to correct simulated PET data for respiratory motion., Methods: PET sinogram data were simulated using SimSET from a 4D pseudo-PET image series created by segmenting MR images acquired over a respiratory cycle. Motion-corrected PET images were produced using post-reconstruction registration (PRR) and motion-compensated image reconstruction (MCIR)., Results: MRI-based motion correction improved PET image quality at the lung-liver and lung-spleen boundaries and in the heart but little improvement was obtained where MRI contrast was low. The root mean square error in SUV units per voxel compared to a motion-free image was reduced from 0.0271 (no motion correction) to 0.0264 (PRR) and 0.0250 (MCIR)., Conclusions: Motion correction using MRI can improve thoracic PET images but there are limitations due to the quality of fast MRI.