Machine learning-based analysis of MR radiomics can help to improve the diagnostic performance of PI-RADS v2 in clinically relevant prostate cancer

OBJECTIVE

To investigate whether machine learning-based analysis of MR radiomics can help improve the performance PI-RADS v2 in clinically relevant prostate cancer (PCa).

METHODS

This IRB-approved study included 54 patients with PCa undergoing multi-parametric (mp) MRI before prostatectomy. Imaging analysis was performed on 54 tumours, 47 normal peripheral (PZ) and 48 normal transitional (TZ) zone based on histological-radiological correlation. Mp-MRI was scored via PI-RADS, and quantified by measuring radiomic features. Predictive model was developed using a novel support vector machine trained with: (i) radiomics, (ii) PI-RADS scores, (iii) radiomics and PI-RADS scores. Paired comparison was made via ROC analysis.

RESULTS

For PCa versus normal TZ, the model trained with radiomics had a significantly higher area under the ROC curve (Az) (0.955 [95% CI 0.923-0.976]) than PI-RADS (Az: 0.878 [0.834-0.914], p < 0.001). The Az between them was insignificant for PCa versus PZ (0.972 [0.945-0.988] vs. 0.940 [0.905-0.965], p = 0.097). When radiomics was added, performance of PI-RADS was significantly improved for PCa versus PZ (Az: 0.983 [0.960-0.995]) and PCa versus TZ (Az: 0.968 [0.940-0.985]).

CONCLUSION

Machine learning analysis of MR radiomics can help improve the performance of PI-RADS in clinically relevant PCa.

KEY POINTS

• Machine-based analysis of MR radiomics outperformed in TZ cancer against PI-RADS. • Adding MR radiomics significantly improved the performance of PI-RADS. • DKI-derived Dapp and Kapp were two strong markers for the diagnosis of PCa.

Intra-individual comparison of (68)Ga-PSMA-11-PET/CT and multi-parametric MR for imaging of primary prostate cancer

Purpose

Multi-parametric magnetic resonance imaging (MP-MRI) is currently the most comprehensive work up for non-invasive primary tumor staging of prostate cancer (PCa). Prostate-specific membrane antigen (PSMA)-Positron emission tomography–computed tomography (PET/CT) is presented to be a highly promising new technique for N- and M-staging in recurrent PCa-patients. The actual investigation analyses the potential of ⁶⁸Ga-PSMA11-PET/CT to assess the extent of primary prostate cancer by intra-individual comparison to MP-MRI.

Methods

In a retrospective study, ten patients with primary PCa underwent MP-MRI and PSMA-PET/CT for initial staging. All tumors were proven histopathological by biopsy. Image analysis was done in a quantitative (SUVmax) and qualitative (blinded read) fashion based on PI-RADS. The PI-RADS schema was then translated into a 3D-matrix and the euclidian distance of this coordinate system was used to quantify the extend of agreement.

Results

Both MP-MRI and PSMA-PET/CT presented a good allocation of the PCa, which was also in concordance to the tumor location validated in eight-segment resolution by biopsy. An Isocontour of 50 % SUVmax in PSMA-PET resulted in visually concordant tumor extension in comparison to MP-MRI (T2w and DWI). For 89.4 % of sections containing a tumor according to MP-MRI, the tumor was also identified in total or near-total agreement (euclidian distance ≤1) by PSMA-PET. Vice versa for 96.8 % of the sections identified as tumor bearing by PSMA-PET the tumor was also found in total or near-total agreement by MP-MRI.

Conclusions

PSMA-PET/CT and MP-MRI correlated well with regard to tumor allocation in patients with a high pre-test probability for large tumors. Further research will be needed to evaluate its value in challenging situation such as prostatitis or after repeated negative biopsies.

Machine learning for prediction of chemoradiation therapy response in rectal cancer using pre-treatment and mid-radiation multi-parametric MRI

PURPOSE

To predict the neoadjuvant chemoradiation therapy (CRT) response in patients with locally advanced rectal cancer (LARC) using radiomics and deep learning based on pre-treatment MRI and a mid-radiation follow-up MRI taken 3-4 weeks after the start of CRT.

METHODS

A total of 51 patients were included, 45 with pre-treatment, 41 with mid-radiation therapy (RT), and 35 with both MRI sets. The multi-parametric MRI protocol included T2, diffusion weighted imaging (DWI) with b-values of 0 and 800 s/mm², and dynamic-contrast-enhanced (DCE) MRI. After completing CRT and surgery, the specimen was examined to determine the pathological response based on the tumor regression grade. The tumor ROI was manually drawn on the post-contrast image and mapped to other sequences. The total tumor volume and mean apparent diffusion coefficient (ADC) were measured. Radiomics using GLCM texture and histogram parameters, and deep learning using a convolutional neural network (CNN), were performed to differentiate pathologic complete response (pCR) vs. non-pCR, and good response (GR) vs. non-GR.

RESULTS

Tumor volume decreased and ADC increased significantly in the mid-RT MRI compared to the pre-treatment MRI. For predicting pCR vs. non-pCR, combining ROI and radiomics features achieved an AUC of 0.80 for pre-treatment, 0.82 for mid-RT, and 0.86 for both MRI together. For predicting GR vs. non-GR, the AUC was 0.91 for pre-treatment, 0.92 for mid-RT, and 0.93 for both MRI together. In deep learning using CNN, combining pre-treatment and mid-RT MRI achieved a higher accuracy compared to using either dataset alone, with AUC of 0.83 for predicting pCR vs. non-pCR.

CONCLUSION

Radiomics based on pre-treatment and early follow-up multi-parametric MRI in LARC patients receiving CRT could extract comprehensive quantitative information to predict final pathologic response.

PROMIS--Prostate MR imaging study: A paired validating cohort study evaluating the role of multi-parametric MRI in men with clinical suspicion of prostate cancer

BACKGROUND

Transrectal ultrasound-guided prostate biopsies are prone to detection errors. Multi-parametric MRI (MP-MRI) may improve the diagnostic pathway.

METHODS

PROMIS is a prospective validating paired-cohort study that meets criteria for level 1 evidence in diagnostic test evaluation. PROMIS will investigate whether multi-parametric (MP)-MRI can discriminate between men with and without clinically-significant prostate cancer who are at risk prior to first biopsy. Up to 714 men will have MP-MRI (index), 10-12 core TRUS-biopsy (standard) and 5mm transperineal template mapping (TPM) biopsies (reference). The conduct and reporting of each test will be blinded to the others.

RESULTS

PROMIS will measure and compare sensitivity, specificity, and positive and negative predictive values of both MP-MRI and TRUS-biopsy against TPM biopsies. The MP-MRI results will be used to determine the proportion of men who could safely avoid biopsy without compromising detection of clinically-significant cancers. For the primary outcome, significant cancer on TPM is defined as Gleason grade >/= 4+3 and/or maximum cancer core length of ≥ 6 mm. PROMIS will also assess inter-observer variability among radiologists among other secondary outcomes. Cost-effectiveness of MP-MRI prior to biopsy will also be evaluated.

CONCLUSIONS

PROMIS will determine whether MP-MRI of the prostate prior to first biopsy improves the detection accuracy of clinically-significant cancer.

Longitudinal MRI-based fusion novel model predicts pathological complete response in breast cancer treated with neoadjuvant chemotherapy: a multicenter, retrospective study

Background

Accurate identification of pCR to neoadjuvant chemotherapy (NAC) is essential for determining appropriate surgery strategy and guiding resection extent in breast cancer. However, a non-invasive tool to predict pCR accurately is lacking. Our study aims to develop ensemble learning models using longitudinal multiparametric MRI to predict pCR in breast cancer.

Methods

From July 2015 to December 2021, we collected pre-NAC and post-NAC multiparametric MRI sequences per patient. We then extracted 14,676 radiomics and 4096 deep learning features and calculated additional delta-value features. In the primary cohort (n = 409), the inter-class correlation coefficient test, U-test, Boruta and the least absolute shrinkage and selection operator regression were used to select the most significant features for each subtype of breast cancer. Five machine learning classifiers were then developed to predict pCR accurately for each subtype. The ensemble learning strategy was used to integrate the single-modality models. The diagnostic performances of models were evaluated in the three external cohorts (n = 343, 170 and 340, respectively).

Findings

A total of 1262 patients with breast cancer from four centers were enrolled in this study, and pCR rates were 10.6% (52/491), 54.3% (323/595) and 37.5% (66/176) in HR+/HER2-, HER2+ and TNBC subtype, respectively. Finally, 20, 15 and 13 features were selected to construct the machine learning models in HR+/HER2-, HER2+ and TNBC subtypes, respectively. The multi-Layer Perception (MLP) yields the best diagnostic performances in all subtypes. For the three subtypes, the stacking model integrating pre-, post- and delta-models yielded the highest AUCs of 0.959, 0.974 and 0.958 in the primary cohort, and AUCs of 0.882-0.908, 0.896-0.929 and 0.837-0.901 in the external validation cohorts, respectively. The stacking model had accuracies of 85.0%-88.9%, sensitivities of 80.0%-86.3%, and specificities of 87.4%-91.5% in the external validation cohorts.

Interpretation

Our study established a novel tool to predict the responses of breast cancer to NAC and achieve excellent performance. The models could help to determine post-NAC surgery strategy for breast cancer.

Funding

This study is supported by grants from the National Natural Science Foundation of China (82171898, 82103093), the Deng Feng project of high-level hospital construction (DFJHBF202109), the Guangdong Basic and Applied Basic Research Foundation (grant number, 2020A1515010346, 2022A1515012277), the Science and Technology Planning Project of Guangzhou City (202002030236), the Beijing Medical Award Foundation (YXJL-2020-0941-0758), and the Beijing Science and Technology Innovation Medical Development Foundation (KC2022-ZZ-0091-5). Funding sources were not involved in the study design, data collection, analysis and interpretation, writing of the report, or decision to submit the article for publication.

Deep transfer learning-based prostate cancer classification using 3 Tesla multi-parametric MRI

PURPOSE

The purpose of the study was to propose a deep transfer learning (DTL)-based model to distinguish indolent from clinically significant prostate cancer (PCa) lesions and to compare the DTL-based model with a deep learning (DL) model without transfer learning and PIRADS v2 score on 3 Tesla multi-parametric MRI (3T mp-MRI) with whole-mount histopathology (WMHP) validation.

METHODS

With IRB approval, 140 patients with 3T mp-MRI and WMHP comprised the study cohort. The DTL-based model was trained on 169 lesions in 110 arbitrarily selected patients and tested on the remaining 47 lesions in 30 patients. We compared the DTL-based model with the same DL model architecture trained from scratch and the classification based on PIRADS v2 score with a threshold of 4 using accuracy, sensitivity, specificity, and area under curve (AUC). Bootstrapping with 2000 resamples was performed to estimate the 95% confidence interval (CI) for AUC.

RESULTS

After training on 169 lesions in 110 patients, the AUC of discriminating indolent from clinically significant PCa lesions of the DTL-based model, DL model without transfer learning and PIRADS v2 score ≥ 4 were 0.726 (CI [0.575, 0.876]), 0.687 (CI [0.532, 0.843]), and 0.711 (CI [0.575, 0.847]), respectively, in the testing set. The DTL-based model achieved higher AUC compared to the DL model without transfer learning and PIRADS v2 score ≥ 4 in discriminating clinically significant lesions in the testing set.

CONCLUSION

The DeLong test indicated that the DTL-based model achieved comparable AUC compared to the classification based on PIRADS v2 score (p = 0.89).

Multi-parametric MRI of rectal cancer - do quantitative functional MR measurements correlate with radiologic and pathologic tumor stages?

PURPOSE

The purpose of this study is two-fold. First, to evaluate, whether functional rectal MRI techniques can be analyzed in a reproducible manner by different readers and second, to assess whether different clinical and pathologic T and N stages can be differentiated by functional MRI measurements.

MATERIALS AND METHODS

54 patients (38 men, 16 female; mean age 63.2 ± 12.2 years) with pathologically proven rectal cancer were included in this retrospective IRB-approved study. All patients were referred for a multi-parametric MRI protocol on a 3 Tesla MR-system, consisting of a high-resolution, axial T2 TSE sequence, DWI and perfusion imaging (plasma flow -s PFTumor) prior to any treatment. Two experienced radiologists evaluated the MRI measurements, blinded to clinical data and outcome. Inter-reader correlation and the association of functional MRI parameters with c- and p-staging were analyzed.

RESULTS

The inter-reader correlation for lymph node (ρ 0.76-0.94; p<0.0002) and primary tumor (ρ 0.78-0.92; p<0.0001) apparent diffusion coefficient and plasma flow (PF) values was good to very good. PFTumor values decreased with cT stage with significant differences identified between cT2 and cT3 tumors (229 versus 107.6 ml/100ml/min; p=0.05). ADCTumor values did not differ significantly. No substantial discrepancies in lymph node ADCLn values or short axis diameter were found among cN1-3 stages, whereas PFLn values were distinct between cN1 versus cN2 stages (p=0.03). In the patients without neoadjuvant RCT no statistically significant differences in the assessed functional parameters on the basis of pathologic stage were found.

CONCLUSION

This study illustrates that ADC as well as MR perfusion values can be analyzed with good interobserver agreement in patients with rectal cancer. Moreover, MR perfusion parameters may allow accurate differentiation of tumor stages. Both findings suggest that functional MRI parameters may help to discriminate T and N stages for clinical decision making.

Risk-based Patient Selection for Magnetic Resonance Imaging-targeted Prostate Biopsy after Negative Transrectal Ultrasound-guided Random Biopsy Avoids Unnecessary Magnetic Resonance Imaging Scans

BACKGROUND

Multiparametric magnetic resonance imaging (mpMRI) is increasingly used in men with suspicion of prostate cancer (PCa) after negative transrectal ultrasound (TRUS)-guided random biopsy. Risk-based patient selection for mpMRI could help to avoid unnecessary mpMRIs.

OBJECTIVE

To study the rate of potentially avoided mpMRIs after negative TRUS-guided random biopsy by risk-based patient selection using the Rotterdam Prostate Cancer Risk Calculator (RPCRC).

DESIGN, SETTING, AND PARTICIPANTS

One hundred and twenty two consecutive men received a mpMRI scan and subsequent MRI-TRUS fusion targeted biopsy in case of suspicious lesion(s) (Prostate Imaging Reporting and Data System ≥ 3) after negative TRUS-guided random biopsy. Men were retrospectively stratified according to the RPCRC biopsy advice to compare targeted biopsy outcomes after risk-based patient selection with standard (prostate specific antigen and/or digital rectal examination-driven) patient selection.

OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS

The rate of potentially avoided mpMRIs by RPCRC-based patient selection in relation to the rate of missed high-grade (Gleason ≥ 3+4) PCa. Receiver operating characteristic curve analysis was performed to determine the area under the curve of the RPCRC for (high-grade) PCa.

RESULTS AND LIMITATIONS

Of the 60 men with a positive biopsy advice, six (10%) had low-grade PCa and 28 (47%) had high-grade PCa in targeted biopsy. Of the 62 men with a negative advice, two (3%) had low-grade PCa and three (5%) had high-grade PCa. Upfront RPCRC-based patient selection would have avoided 62 (51%) of 122 mpMRIs and two (25%) of eight low-grade PCa diagnoses, missing three (10%) of 31 high-grade PCa. The area under the curve of the RPCRC for PCa and high-grade PCa was respectively 0.76 (95% confidence interval 0.67-0.85) and 0.84 (95% confidence interval 0.76-0.93).

CONCLUSIONS

Risk-based patient selection with the RPCRC can avoid half of mpMRIs after a negative prostate specific antigen and/or digital rectal examination-driven TRUS-guided random biopsy. Further improvement in risk-based patient selection for mpMRI could be made by adjusting the RPCRC for MRI-targeted biopsy outcome prediction.

PATIENT SUMMARY

The suspicion of prostate cancer remains in many men after a negative ultrasound-guided prostate biopsy. These men increasingly receive an often unnecessary magnetic resonance imaging (MRI) scan. We found that patient selection for MRI based on the Rotterdam Prostate Cancer Risk Calculator biopsy advice could avoid half of the MRIs.

DCE MRI of prostate cancer

DCE MRI is an established component of multi-parametric MRI of the prostate. The sequence highlights the vascularization of cancerous lesions, allowing readers to corroborate suspicious findings on T2W and DW MRI and to note subtle lesions not visible on the other sequences. In this article, we review the technical aspects, methods of evaluation, limitations, and future perspectives of DCE MRI.

Volume of high-risk intratumoral subregions at multi-parametric MR imaging predicts overall survival and complements molecular analysis of glioblastoma

OBJECTIVE

To develop and validate a volume-based, quantitative imaging marker by integrating multi-parametric MR images for predicting glioblastoma survival, and to investigate its relationship and synergy with molecular characteristics.

METHODS

We retrospectively analysed 108 patients with primary glioblastoma. The discovery cohort consisted of 62 patients from the cancer genome atlas (TCGA). Another 46 patients comprising 30 from TCGA and 16 internally were used for independent validation. Based on integrated analyses of T1-weighted contrast-enhanced (T1-c) and diffusion-weighted MR images, we identified an intratumoral subregion with both high T1-c and low ADC, and accordingly defined a high-risk volume (HRV). We evaluated its prognostic value and biological significance with genomic data.

RESULTS

On both discovery and validation cohorts, HRV predicted overall survival (OS) (concordance index: 0.642 and 0.653, P < 0.001 and P = 0.038, respectively). HRV stratified patients within the proneural molecular subtype (log-rank P = 0.040, hazard ratio = 2.787). We observed different OS among patients depending on their MGMT methylation status and HRV (log-rank P = 0.011). Patients with unmethylated MGMT and high HRV had significantly shorter survival (median survival: 9.3 vs. 18.4 months, log-rank P = 0.002).

CONCLUSION

Volume of the high-risk intratumoral subregion identified on multi-parametric MRI predicts glioblastoma survival, and may provide complementary value to genomic information.

KEY POINTS

• High-risk volume (HRV) defined on multi-parametric MRI predicted GBM survival. • The proneural molecular subtype tended to harbour smaller HRV than other subtypes. • Patients with unmethylated MGMT and high HRV had significantly shorter survival. • HRV complements genomic information in predicting GBM survival.

The PRIMARY Score: Using intra-prostatic PSMA PET/CT patterns to optimise prostate cancer diagnosis

Background: Multi-parametric magnetic resonance imaging (mpMRI) is validated for the diagnosis of clinically significant prostate cancer (csPCa). ⁶⁸Ga-PSMA -11 PET/CT (PSMA-PET/CT) combined with mpMRI has improved negative predictive value over mpMRI alone for csPCa. The aim of this post-hoc analysis of the PRIMARY study was to evaluate the clinical significance of patterns of intra-prostatic PSMA activity, proposing a 5- point PRIMARY score to optimise accuracy of PSMA-PET/CT for csPCa in a low prevalence population. Methods: The PRIMARY trial is a prospective multi-centre phase II imaging trial that enrolled biopsy-naïve men with suspected PCa, no prior biopsy, recent mpMRI (6 months) and planned for prostate biopsy. 291 men underwent mpMRI, PSMA-PET/CT and systematic +/- targeted biopsy. The mpMRI was read separately using PI-RADS (V2). PSMA-PET/CT (pelvic only) was acquired a minimum 60 minutes post injection. PSMA-PET/CT was centrally read for pattern (diffuse transition zone (TZ), symmetrical central zone (CZ), focal TZ or peripheral zone (PZ), and intensity (SUV_max). In this post-hoc analysis, a 5-level PRIMARY score was assigned based on analysis of the central read: 1. No pattern, 2. Diffuse TZ or CZ (no focal), 3. Focal TZ, 4. Focal PZ or 5. SUV_max ≥ 12. Two further readers independently assigned a PRIMARY score to 118 scans for inter-rater agreement. Associations between PRIMARY score and csPCa (ISUP≥2) were evaluated. Results: Of 291 men enrolled, 162 (56%) had csPCa. PRIMARY score-1 was present in 16% (47), score-2 in 19% (55), score-3 in 10% (29), score-4 in 40% (117) and score-5 in 15% (43). The proportion of patients with csPCa and PRIMARY score 1 to 5 was 8.5% (4/47), 27% (15/55), 38% (11/29), 76% (89/117) and 100% (43/43) respectively. Sensitivity, specificity, PPV and NPV for PRIMARY score 1,2 (low-risk patterns) vs PRIMARY score 3-5 (high-risk patterns) was 88%, 64%, 76% and 81%, compared to 83%, 53%, 69% and 72% for PI-RADS (2 vs 3-5) on mpMRI. The inter-rater agreements for PRIMARY score 1,2 vs. PRIMARY score 3-5 was 0.76 (CI: 0.64-0.88) and 0.64 (CI: 0.49-0.78). Conclusion: A PRIMARY score incorporating intra-prostatic pattern and intensity on PSMA-PET/CT shows potential with high diagnostic accuracy for csPCa. Further validation is warranted prior to implementation.

Fully-integrated framework for the segmentation and registration of the spinal cord white and gray matter

The spinal cord white and gray matter can be affected by various pathologies such as multiple sclerosis, amyotrophic lateral sclerosis or trauma. Being able to precisely segment the white and gray matter could help with MR image analysis and hence be useful in further understanding these pathologies, and helping with diagnosis/prognosis and drug development. Up to date, white/gray matter segmentation has mostly been done manually, which is time consuming, induces a bias related to the rater and prevents large-scale multi-center studies. Recently, few methods have been proposed to automatically segment the spinal cord white and gray matter. However, no single method exists that combines the following criteria: (i) fully automatic, (ii) works on various MRI contrasts, (iii) robust towards pathology and (iv) freely available and open source. In this study we propose a multi-atlas based method for the segmentation of the spinal cord white and gray matter that addresses the previous limitations. Moreover, to study the spinal cord morphology, atlas-based approaches are increasingly used. These approaches rely on the registration of a spinal cord template to an MR image, however the registration usually doesn't take into account the spinal cord internal structure and thus lacks accuracy. In this study, we propose a new template registration framework that integrates the white and gray matter segmentation to account for the specific gray matter shape of each individual subject. Validation of segmentation was performed in 24 healthy subjects using T₂^*-weighted images, in 8 healthy subjects using diffusion weighted images (exhibiting inverted white-to-gray matter contrast compared to T₂*-weighted), and in 5 patients with spinal cord injury. The template registration was validated in 24 subjects using T₂*-weighted data. Results of automatic segmentation on T₂*-weighted images was in close correspondence with the manual segmentation (Dice coefficient in the white/gray matter of 0.91/0.71 respectively). Similarly, good results were obtained in data with inverted contrast (diffusion-weighted image) and in patients. When compared to the classical template registration framework, the proposed framework that accounts for gray matter shape significantly improved the quality of the registration (comparing Dice coefficient in gray matter: p=9.5×10^-6). While further validation is needed to show the benefits of the new registration framework in large cohorts and in a variety of patients, this study provides a fully-integrated tool for quantitative assessment of white/gray matter morphometry and template-based analysis. All the proposed methods are implemented in the Spinal Cord Toolbox (SCT), an open-source software for processing spinal cord multi-parametric MRI data.

MPCaD: a multi-scale radiomics-driven framework for automated prostate cancer localization and detection

Background

Quantitative radiomic features provide a plethora of minable data extracted from multi-parametric magnetic resonance imaging (MP-MRI) which can be used for accurate detection and localization of prostate cancer. While most cancer detection algorithms utilize either voxel-based or region-based feature models, the complexity of prostate tumour phenotype in MP-MRI requires a more sophisticated framework to better leverage available data and exploit a priori knowledge in the field.

Methods

In this paper, we present MPCaD, a novel Multi-scale radiomics-driven framework for Prostate Cancer Detection and localization which leverages radiomic feature models at different scales as well as incorporates a priori knowledge of the field. Tumour candidate localization is first performed using a statistical texture distinctiveness strategy that leverages a voxel-resolution feature model to localize tumour candidate regions. Tumour region classification via a region-resolution feature model is then performed to identify tumour regions. Both voxel-resolution and region-resolution feature models are built upon and extracted from six different MP-MRI modalities. Finally, a conditional random field framework that is driven by voxel-resolution relative ADC features is used to further refine the localization of the tumour regions in the peripheral zone to improve the accuracy of the results.

Results

The proposed framework is evaluated using clinical prostate MP-MRI data from 30 patients, and results demonstrate that the proposed framework exhibits enhanced separability of cancerous and healthy tissue, as well as outperforms individual quantitative radiomics models for prostate cancer detection.

Conclusion

Quantitative radiomic features extracted from MP-MRI of prostate can be utilized to detect and localize prostate cancer.

Comparison of unsupervised classification methods for brain tumor segmentation using multi-parametric MRI

Tumor segmentation is a particularly challenging task in high-grade gliomas (HGGs), as they are among the most heterogeneous tumors in oncology. An accurate delineation of the lesion and its main subcomponents contributes to optimal treatment planning, prognosis and follow-up. Conventional MRI (cMRI) is the imaging modality of choice for manual segmentation, and is also considered in the vast majority of automated segmentation studies. Advanced MRI modalities such as perfusion-weighted imaging (PWI), diffusion-weighted imaging (DWI) and magnetic resonance spectroscopic imaging (MRSI) have already shown their added value in tumor tissue characterization, hence there have been recent suggestions of combining different MRI modalities into a multi-parametric MRI (MP-MRI) approach for brain tumor segmentation. In this paper, we compare the performance of several unsupervised classification methods for HGG segmentation based on MP-MRI data including cMRI, DWI, MRSI and PWI. Two independent MP-MRI datasets with a different acquisition protocol were available from different hospitals. We demonstrate that a hierarchical non-negative matrix factorization variant which was previously introduced for MP-MRI tumor segmentation gives the best performance in terms of mean Dice-scores for the pathologic tissue classes on both datasets.

•

Unsupervised classification algorithms are applied for brain tumor segmentation on multi-parametric MRI datasets.

•

Reported mean Dice-scores are in the range of state-of-the-art segmentation algorithms.

•

Hierarchical NMF obtained the best segmentation results in terms of mean Dice-scores for most of the tissue classes.

Multi-parametric MRI-guided focal tumor boost using HDR prostate brachytherapy: a feasibility study

PURPOSE

This study investigates the feasibility of delivering focal boost dose to tumor regions, identified with multi-parametric MRI, in high-dose-rate prostate brachytherapy.

METHODS AND MATERIALS

T2-weighted, diffusion-weighted, and dynamic-contrast-enhanced MRI were acquired the day before treatment and analyzed retrospectively for 15 patients. Twelve patients had hormone therapy before the MRI scan. The tumor was delineated on MRI by a radiologist and registered to treatment planning transrectal ultrasound images. A margin based on analysis of delineation and registration uncertainties was applied to create a focal boost planning target volume (F-PTV). Delivered treatment plans were compared with focal boost plans optimized to increase F-PTV dose as much as allowed by urethral and rectal dose constraints.

RESULTS

Tumors were delineated in all patients with volumes 0.4-23.0cc. The margin for tumor delineation and image registration uncertainties was estimated to be 4.5 mm. For F-PTV, the focal boost treatment plans increased median D90 from 17.6 to 20.9 Gy and median V150 from 27.3% to 75.9%.

CONCLUSIONS

MRI-guided high-dose-rate prostate brachytherapy focal tumor boost is feasible-tumor regions can be identified even after hormone therapy, and focal boost dose can be delivered without violating urethral and rectal dose constraints.

Multi-parametric MRI in the early prediction of response to neo-adjuvant chemotherapy in breast cancer: Value of non-modelled parameters

OBJECTIVE

To prospectively evaluate individual functional MRI metrics for the early prediction of pathological complete response (pCR) to neo-adjuvant chemotherapy (NAC) in breast cancer.

MATERIALS AND METHODS

Thirty-two women (median age 52 years; range 32-71 years) with biopsy proven breast cancer due to receive neo-adjuvant anthracycline and/or taxane-based chemotherapy were prospectively recruited following local research ethics committee approval and written informed consent. Breast MRI was performed prior to and after two cycles of NAC and pCR was assessed after surgery. The enhancement fraction (EF), tumour volume, initial area under the gadolinium curve (IAUGC), pharmacokinetic parameters (K(trans), kep and ve), the apparent diffusion coefficient (ADC) and R2* values, along with the percentage change in these parameters after two cycles were evaluated according to pCR status using an independent samples t-test. The area under the receiver operating characteristics curve (AUC) was calculated for each parameter. Linear discriminant analysis (LDA) determined the most important parameter in predicting pCR.

RESULTS

A reduction in the EF (-41% ± 38%) and tumour volume (-80% ± 25%) after 2 cycles of NAC were significantly greater in those achieving pCR (p=0.025, p=0.011 respectively). A reduction in the EF of 7% after 2 cycles of NAC identified those more likely to achieve pCR (AUC 0.76). AUC changes in other parameters were tumour volume (0.77), IAUGC (0.64), K(trans) (0.60), kep (0.68), ve (0.58), ADC (0.69) and R2* (0.41).

CONCLUSION

In a multi-parametric MRI model, the decrease in a non-model based vascular parameter the enhancement fraction as well as the tumour volume are the most important early predictors of pCR in breast cancer.

MRI in acute muscle tears in athletes: can quantitative T2 and DTI predict return to play better than visual assessment?

OBJECTIVES

To assess the ability of quantitative T2, diffusion tensor imaging (DTI) and radiologist's scores to detect muscle changes following acute muscle tear in soccer and rugby players. To assess the ability of these parameters to predict return to play times.

METHODS

In this prospective, longitudinal study, 13 male athletes (age 19 to 34 years; mean 25 years) underwent MRI within 1 week of suffering acute muscle tear. Imaging included measurements of T2 and DTI parameters. Images were also assessed using modified Peetrons and British athletics muscle injury classification (BAMIC) scores. Participants returned for a second scan within 1 week of being determined fit to return to play. MRI measurements were compared between visits. Pearson's correlation between visit 1 measurements and return to play times was assessed.

RESULTS

There were significant differences between visits in BAMIC scores (Z = - 2.088; p = 0.037), modified Peetrons (Z = - 2.530; p = 0.011) and quantitative MRI measurements; T2, 13.12 ms (95% CI, 4.82 ms, 21.42 ms; p = 0.01); mean diffusivity (0.22 (0.04, 0.39); p = 0.02) and fractional anisotropy (0.07 (0.01, 0.14); p = 0.03). BAMIC scores showed a significant correlation with return to play time (Rs = 0.64; p = 0.02), but modified Peetrons scores and quantitative parameters did not.

CONCLUSIONS

T2 and DTI measurements in muscle can detect changes due to healing following muscle tear. Although BAMIC scores correlated well with return to play times, in this small study, quantitative MRI values did not, suggesting that T2 and DTI measurements are inferior predictors of return to play time compared with visual scoring.

KEY POINTS

• Muscle changes following acute muscle tear can be measured using T2 and diffusion measurements on MRI. • Measurements of T2 and diffusion using MRI are not as good as a radiologist's visual report at predicting return to play time after acute muscle tear.

Multi-parametric quantitative in vivo spinal cord MRI with unified signal readout and image denoising

Multi-parametric quantitative MRI (qMRI) of the spinal cord is a promising non-invasive tool to probe early microstructural damage in neurological disorders. It is usually performed in vivo by combining acquisitions with multiple signal readouts, which exhibit different thermal noise levels, geometrical distortions and susceptibility to physiological noise. This ultimately hinders joint multi-contrast modelling and makes the geometric correspondence of parametric maps challenging. We propose an approach to overcome these limitations, by implementing state-of-the-art microstructural MRI of the spinal cord with a unified signal readout in vivo (i.e. with matched spatial encoding parameters across a range of imaging contrasts). We base our acquisition on single-shot echo planar imaging with reduced field-of-view, and obtain data from two different vendors (vendor 1: Philips Achieva; vendor 2: Siemens Prisma). Importantly, the unified acquisition allows us to compare signal and noise across contrasts, thus enabling overall quality enhancement via multi-contrast image denoising methods. As a proof-of-concept, here we provide a demonstration with one such method, known as Marchenko-Pastur (MP) Principal Component Analysis (PCA) denoising. MP-PCA is a singular value (SV) decomposition truncation approach that relies on redundant acquisitions, i.e. such that the number of measurements is large compared to the number of components that are maintained in the truncated SV decomposition. Here we used in vivo and synthetic data to test whether a unified readout enables more efficient MP-PCA denoising of less redundant acquisitions, since these can be denoised jointly with more redundant ones. We demonstrate that a unified readout provides robust multi-parametric maps, including diffusion and kurtosis tensors from diffusion MRI, myelin metrics from two-pool magnetisation transfer, and T1 and T2 from relaxometry. Moreover, we show that MP-PCA improves the quality of our multi-contrast acquisitions, since it reduces the coefficient of variation (i.e. variability) by up to 17% for mean kurtosis, 8% for bound pool fraction (myelin-sensitive), and 13% for T1, while enabling more efficient denoising of modalities limited in redundancy (e.g. relaxometry). In conclusion, multi-parametric spinal cord qMRI with unified readout is feasible and provides robust microstructural metrics with matched resolution and distortions, whose quality benefits from multi-contrast denoising methods such as MP-PCA.

Does focal incidental 18F-FDG PET/CT uptake in the prostate have significance?

PURPOSE

(18)F-FDG PET/CT is used to characterize many malignancies, but is not recommended for localized prostate cancer. This study explores the value of multi-parametric MRI (mpMRI) in characterizing incidental prostate (18)F-FDG uptake.

METHODS

Thirty-one patients who underwent (18)F-FDG PET/CT for reasons unrelated to prostate cancer and prostate mpMRI were eligible for this retrospective study. The mpMRI included T2-weighted (T2W), dynamic contrast enhancement (DCE), apparent diffusion coefficient (ADC), and MR spectroscopy (MRS) sequences. Fourteen patients were excluded (n = 8 insufficient histopathology, n = 6 radical prostatectomy before PET), and final analysis included 17 patients. A nuclear medicine physician, blinded to clinicopathologic findings, identified suspicious areas and maximum standardized uptake values (SUV max) on (18)F-FDG PET/CT. Sector-based imaging findings were correlated with annotated histopathology from whole-mount or MRI/transrectal ultrasound fusion biopsy samples. Positive predictive values (PPVs) were estimated using generalized estimating equations with logit link. Results were evaluated with Kruskal-Wallis and Dunn's multiple comparisons tests.

RESULTS

The PPV of (18)F-FDG PET alone in detecting prostate cancer was 0.65. Combining (18)F-FDG PET as a base parameter with mpMRI (T2W, DCE, ADC, and MRS) increased the PPV to 0.82, 0.83, 0.83, and 0.94, respectively. All benign lesions had SUV max < 6. Malignant lesions had higher SUV max values that correlated with Gleason scores. There was a significant difference in SUV max per prostate between the Gleason ≥ 4 + 5 and benign categories (p = 0.03).

CONCLUSIONS

Focal incidental prostate (18)F-FDG uptake has low clinical utility alone, but regions of uptake may harbor high-grade prostate cancer, especially if SUV max > 6. Using mpMRI to further evaluate incidental (18)F-FDG uptake aids the diagnosis of prostate cancer.

GAS: A genetic atlas selection strategy in multi-atlas segmentation framework

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.

Ruling out clinically significant prostate cancer with negative multi-parametric MRI

PURPOSE

To evaluate the negative predictive value (NPV) of a negative prostate multi-parametric magnetic resonance imaging (mpMRI) in ruling out clinically significant prostate upon 12-core systematic biopsy.

METHODS

We retrospectively reviewed 114 men evaluated at our institution who underwent systematic 12-core biopsy within 1 year of a negative prostate mpMRI. Clinicopathologic features were evaluated and NPV was calculated for detection of clinically significant (Gleason ≥ 7) cancer. Regression analysis was performed to identify clinical predictors of biopsy outcome.

RESULTS

Overall, 88 (77.2%) patients in our cohort had no cancer detected upon biopsy. The highest pathologic grade was Gleason 6 (3 + 3) in 22 (19.3%) patients, and Gleason ≥ 7 in 4 (3.6%) patients. NPV for detecting Gleason ≥ 7 cancer was 96.5% (95% CI 93.1-99.9%) in the entire negative MRI cohort, 100% in those who were prostate biopsy naïve (n = 20), 100% in those with a prior negative biopsy (n = 53), and 90% in those who have had a previous positive biopsy and on active surveillance (n = 41). Regression analysis identified no predictors of significant cancer in our cohort.

CONCLUSION

In our cohort of men with no lesions detected on prostate mpMRI, we found very low rates of clinically significant cancer on systematic 12-core biopsy. In the few patients who diagnosed with prostate cancer, the majority had low-risk disease and could remain on active surveillance. Although validation studies and greater sample size is needed before clinical recommendations can be made, our data suggest patients with negative mpMRI evaluated by experienced radiologists may avoid unnecessary prostate biopsy and potential overtreatment.

Convolutional neural networks for head and neck tumor segmentation on 7-channel multiparametric MRI: a leave-one-out analysis

Background

Automatic tumor segmentation based on Convolutional Neural Networks (CNNs) has shown to be a valuable tool in treatment planning and clinical decision making. We investigate the influence of 7 MRI input channels of a CNN with respect to the segmentation performance of head&neck cancer.

Methods

Head&neck cancer patients underwent multi-parametric MRI including T2w, pre- and post-contrast T1w, T2*, perfusion (k_trans, v_e) and diffusion (ADC) measurements at 3 time points before and during radiochemotherapy. The 7 different MRI contrasts (input channels) and manually defined gross tumor volumes (primary tumor and lymph node metastases) were used to train CNNs for lesion segmentation. A reference CNN with all input channels was compared to individually trained CNNs where one of the input channels was left out to identify which MRI contrast contributes the most to the tumor segmentation task. A statistical analysis was employed to account for random fluctuations in the segmentation performance.

Results

The CNN segmentation performance scored up to a Dice similarity coefficient (DSC) of 0.65. The network trained without T2* data generally yielded the worst results, with ΔDSC_GTV-T = 5.7% for primary tumor and ΔDSC_GTV-Ln = 5.8% for lymph node metastases compared to the network containing all input channels. Overall, the ADC input channel showed the least impact on segmentation performance, with ΔDSC_GTV-T = 2.4% for primary tumor and ΔDSC_GTV-Ln = 2.2% respectively.

Conclusions

We developed a method to reduce overall scan times in MRI protocols by prioritizing those sequences that add most unique information for the task of automatic tumor segmentation. The optimized CNNs could be used to aid in the definition of the GTVs in radiotherapy planning, and the faster imaging protocols will reduce patient scan times which can increase patient compliance.

Trial registration

The trial was registered retrospectively at the German Register for Clinical Studies (DRKS) under register number DRKS00003830 on August 20th, 2015.

Evaluating the size criterion for PI-RADSv2 category 5 upgrade: is 15 mm the best threshold?

PURPOSE

The purpose of the study was to determine if the ≥ 15 mm threshold currently used to define PIRADS 5 lesions is the optimal size threshold for predicting high likelihood of clinically significant (CS) cancers.

MATERIALS

Three hundred and fifty-eight lesions that may be changed from category 4 to 5 or vice versa on the basis of the size criterion (category 4: n = 288, category 5: n = 70) from 255 patients were evaluated. Kendall's tau-b statistic accounting for inter-lesion correlation, generalized estimation equation logistic regression, and receiver operating curve analysis evaluated two lesion size-metrics (lesion diameter and relative lesion diameter-defined as lesion diameter/prostate volume) for ability to identify CS (Gleason grade ≥ 3 + 4) cancer at targeted biopsy. Optimal cut-points were identified using the Youden index. Analyses were performed for the whole prostate (WP) and zone-specific sub-cohorts of lesions in the peripheral and transition zones (PZ and TZ).

RESULTS

Lesion diameter showed a modest correlation with Gleason grade (WP: τB = 0.21, p < 0.0001; PZ: τB = 0.13, p = 0.02; TZ: τB = 0.32, p = 0.001), and association with CS cancer detection (WP: AUC = 0.63, PZ: AUC = 0.59, TZ: AUC = 0.74). Empirically derived thresholds (WP: 14 mm, PZ: 13 mm, TZ: 16 mm) performed similarly to the current ≥ 15 mm standard. Lesion relative lesion diameter improved identification of CS cancers compared to lesion diameter alone (WP: τB = 0.30, PZ: τB = 0.24, TZ: τB = 0.42, all p < 0.0001). AUC also improved for WP and PZ lesions (WP: AUC = 0.70, PZ: AUC = 0.68, and TZ: AUC = 0.74).

CONCLUSIONS

The current ≥ 15 mm diameter threshold is a reasonable delineator of PI-RADS category 4 and category 5 lesions in the absence of extraprostatic extension to predict CS cancers. Additionally, relative lesion diameter can improve identification of CS cancers and may serve as another option for distinguishing category 4 and 5 lesions.

Prostate cancer detection using residual networks

PURPOSE

To automatically identify regions where prostate cancer is suspected on multi-parametric magnetic resonance images (mp-MRI).

METHODS

A residual network was implemented based on segmentations from an expert radiologist on T2-weighted, apparent diffusion coefficient map, and high b-value diffusion-weighted images. Mp-MRIs from 346 patients were used in this study.

RESULTS

The residual network achieved a hit or miss accuracy of 93% for lesion detection, with an average Jaccard score of 71% that compared the agreement between network and radiologist segmentations.

CONCLUSION

This paper demonstrated the ability for residual networks to learn features for prostate lesion segmentation.

Personalized pathology maps to quantify diffuse and focal brain damage

Background and objectives

Quantitative MRI (qMRI) permits the quantification of brain changes compatible with inflammation, degeneration and repair in multiple sclerosis (MS) patients. In this study, we propose a new method to provide personalized maps of tissue alterations and longitudinal brain changes based on different qMRI metrics, which provide complementary information about brain pathology.

Methods

We performed baseline and two-years follow-up on (i) 13 relapsing-remitting MS patients and (ii) four healthy controls. A group consisting of up to 65 healthy controls was used to compute the reference distribution of qMRI metrics in healthy tissue. All subjects underwent 3T MRI examinations including T1, T2, T2* relaxation and Magnetization Transfer Ratio (MTR) imaging. We used a recent partial volume estimation algorithm to estimate the concentration of different brain tissue types on T1 maps; then, we computed a deviation map (z-score map) for each contrast at both time-points. Finally, we subtracted those deviation maps only for voxels showing a significant difference with healthy tissue in one of the time points, to obtain a difference map for each subject.

Results and conclusion

Control subjects did not show any significant z-score deviations or longitudinal z-score changes. On the other hand, MS patients showed brain regions with cross-sectional and longitudinal concomitant increase in T1, T2, T2* z-scores and decrease of MTR z-scores, suggesting brain tissue degeneration/loss. In the lesion periphery, we observed areas with cross-sectional and longitudinal decreased T1/T2 and slight decrease in T2* most likely related to iron accumulation. Moreover, we measured longitudinal decrease in T1, T2 - and to a lesser extent in T2* - as well as a concomitant increase in MTR, suggesting remyelination/repair.

In summary, we have developed a method that provides whole-brain personalized maps of cross-sectional and longitudinal changes in MS patients, which are computed in patient space. These maps may open new perspectives to complement and support radiological evaluation of brain damage for a given patient.