Cluster analysis of dynamic contrast enhanced MRI reveals tumor subregions related to locoregional relapse for cervical cancer patients.

BACKGROUND: Solid tumors are known to be spatially heterogeneous. Detection of treatment-resistant tumor regions can improve clinical outcome, by enabling implementation of strategies targeting such regions. In this study, K-means clustering was used to group voxels in dynamic contrast enhanced magnetic resonance images (DCE-MRI) of cervical cancers. The aim was to identify clusters reflecting treatment resistance that could be used for targeted radiotherapy with a dose-painting approach. MATERIAL AND METHODS: Eighty-one patients with locally advanced cervical cancer underwent DCE-MRI prior to chemoradiotherapy. The resulting image time series were fitted to two pharmacokinetic models, the Tofts model (yielding parameters Ktrans and νe) and the Brix model (ABrix, kep and kel). K-means clustering was used to group similar voxels based on either the pharmacokinetic parameter maps or the relative signal increase (RSI) time series. The associations between voxel clusters and treatment outcome (measured as locoregional control) were evaluated using the volume fraction or the spatial distribution of each cluster. RESULTS: One voxel cluster based on the RSI time series was significantly related to locoregional control (adjusted p-value 0.048). This cluster consisted of low-enhancing voxels. We found that tumors with poor prognosis had this RSI-based cluster gathered into few patches, making this cluster a potential candidate for targeted radiotherapy. None of the voxels clusters based on Tofts or Brix parameter maps were significantly related to treatment outcome. CONCLUSION: We identified one group of tumor voxels significantly associated with locoregional relapse that could potentially be used for dose painting. This tumor voxel cluster was identified using the raw MRI time series rather than the pharmacokinetic maps.

Classification of dynamic contrast enhanced MR images of cervical cancers using texture analysis and support vector machines.

Dynamic contrast enhanced MRI (DCE-MRI) provides insight into the vascular properties of tissue. Pharmacokinetic models may be fitted to DCE-MRI uptake patterns, enabling biologically relevant interpretations. The aim of our study was to determine whether treatment outcome for 81 patients with locally advanced cervical cancer could be predicted from parameters of the Brix pharmacokinetic model derived from pre-chemoradiotherapy DCE-MRI. First-order statistical features of the Brix parameters were used. In addition, texture analysis of Brix parameter maps was done by constructing gray level co-occurrence matrices (GLCM) from the maps. Clinical factors and first- and second-order features were used as explanatory variables for support vector machine (SVM) classification, with treatment outcome as response. Classification models were validated using leave-one-out cross-model validation. A random value permutation test was used to evaluate model significance. Features derived from first-order statistics could not discriminate between cured and relapsed patients (specificity 0%-20%, p-values close to unity). However, second-order GLCM features could significantly predict treatment outcome with accuracies (~70%) similar to the clinical factors tumor volume and stage (69%). The results indicate that the spatial relations within the tumor, quantified by texture features, were more suitable for outcome prediction than first-order features.

Pharmacokinetic parameters derived from dynamic contrast enhanced MRI of cervical cancers predict chemoradiotherapy outcome.

PURPOSE: To assess the prognostic value of pharmacokinetic parameters derived from pre-chemoradiotherapy dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) of cervical cancer patients. MATERIALS AND METHODS: Seventy-eight patients with locally advanced cervical cancer underwent DCE-MRI with Gd-DTPA before chemoradiotherapy. The pharmacokinetic Brix and Tofts models were fitted to contrast enhancement curves in all tumor voxels, providing histograms of several pharmacokinetic parameters (Brix: A(Brix), k(ep), k(el), Tofts: K(trans), ν(e)). A percentile screening approach including log-rank survival tests was undertaken to identify the clinically most relevant part of the intratumoral parameter distribution. Clinical endpoints were progression-free survival (PFS) and locoregional control (LRC). Multivariate analysis including FIGO stage and tumor volume was used to assess the prognostic significance of the imaging parameters. RESULTS: A(Brix), k(el), and K(trans) were significantly (P<0.05) positively associated with both clinical LRC and PFS, while ν(e) was significantly positively correlated with PFS only. k(ep) showed no association with any endpoint. A(Brix) was positively correlated with K(trans) and ν(e), and showed the strongest association with endpoint in the log-rank testing. k(el) and K(trans) were independent prognostic factors in multivariate analysis with LRC as endpoint. CONCLUSIONS: Parameters estimated by pharmacokinetic analysis of DCE-MR images obtained prior to chemoradiotherapy may be used for identifying patients at risk of treatment failure.

Dynamic contrast-enhanced MRI of cervical cancers: temporal percentile screening of contrast enhancement identifies parameters for prediction of chemoradioresistance.

PURPOSE: To systematically screen the tumor contrast enhancement of locally advanced cervical cancers to assess the prognostic value of two descriptive parameters derived from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). METHODS AND MATERIALS: This study included a prospectively collected cohort of 81 patients who underwent DCE-MRI with gadopentetate dimeglumine before chemoradiotherapy. The following descriptive DCE-MRI parameters were extracted voxel by voxel and presented as histograms for each time point in the dynamic series: normalized relative signal increase (nRSI) and normalized area under the curve (nAUC). The first to 100th percentiles of the histograms were included in a log-rank survival test, resulting in p value and relative risk maps of all percentile-time intervals for each DCE-MRI parameter. The maps were used to evaluate the robustness of the individual percentile-time pairs and to construct prognostic parameters. Clinical endpoints were locoregional control and progression-free survival. The study was approved by the institutional ethics committee. RESULTS: The p value maps of nRSI and nAUC showed a large continuous region of percentile-time pairs that were significantly associated with locoregional control (p < 0.05). These parameters had prognostic impact independent of tumor stage, volume, and lymph node status on multivariate analysis. Only a small percentile-time interval of nRSI was associated with progression-free survival. CONCLUSIONS: The percentile-time screening identified DCE-MRI parameters that predict long-term locoregional control after chemoradiotherapy of cervical cancer.

Pharmacokinetic analysis and k-means clustering of DCEMR images for radiotherapy outcome prediction of advanced cervical cancers.

INTRODUCTION: Pharmacokinetic analysis of dynamic contrast enhanced magnetic resonance images (DCEMRI) allows for quantitative characterization of vascular properties of tumors. The aim of this study is twofold, first to determine if tumor regions with similar vascularization could be labeled by clustering methods, second to determine if the identified regions can be associated with local cancer relapse. MATERIALS AND METHODS: Eighty-one patients with locally advanced cervical cancer treated with chemoradiotherapy underwent DCEMRI with Gd-DTPA prior to external beam radiotherapy. The median follow-up time after treatment was four years, in which nine patients had primary tumor relapse. By fitting a pharmacokinetic two-compartment model function to the temporal contrast enhancement in the tumor, two pharmacokinetic parameters, K(trans) and ύ(e), were estimated voxel by voxel from the DCEMR-images. Intratumoral regions with similar vascularization were identified by k-means clustering of the two pharmacokinetic parameter estimates over all patients. The volume fraction of each cluster was used to evaluate the prognostic value of the clusters. RESULTS: Three clusters provided a sufficient reduction of the cluster variance to label different vascular properties within the tumors. The corresponding median volume fraction of each cluster was 38%, 46% and 10%. The second cluster was significantly associated with primary tumor control in a log-rank survival test (p-value: 0.042), showing a decreased risk of treatment failure for patients with high volume fraction of voxels. CONCLUSIONS: Intratumoral regions showing similar vascular properties could successfully be labeled in three distinct clusters and the volume fraction of one cluster region was associated with primary tumor control.

DCEMRI of spontaneous canine tumors during fractionated radiotherapy: a pharmacokinetic analysis.

PURPOSE: To estimate pharmacokinetic parameters from dynamic contrast-enhanced magnetic resonance (DCEMR) images of spontaneous canine tumors taken during the course of fractionated radiotherapy, and to quantify treatment-induced changes in these parameters. MATERIALS AND METHODS: Six dogs with tumors in the oral or nasal cavity received fractionated conformal radiotherapy with 54 Gy given in 18 fractions. T(1)-weighted DCEMR imaging was performed prior to each treatment fraction. Time-intensity curves in the tumor were extracted voxel-by-voxel, and were fitted to the Brix pharmacokinetic model. The dependence of the pharmacokinetic parameters on the accumulated radiation dose was calculated. RESULTS: The Brix model reproduced the time-intensity curves well. A reduction in the k(ep) parameter with accumulated radiation dose was found for five (three significant) out of six cases, while the results for the A parameter were less consistent. Both pre-treatment k(ep) and the change in k(ep) with accumulated dose correlated significantly with tumor regression. CONCLUSIONS: Pharmacokinetic parameters derived from DCEMR images taken during fractionated radiotherapy may predict response to radiotherapy. This may potentially impact on patient stratification and monitoring of treatment response for image-guided treatment strategies.