Gaussian mixture model-based classification of dynamic contrast enhanced MRI data for identifying diverse tumor microenvironments: preliminary results

Evaluation of functional and metabolic tumor volume using voxel-wise analysis in childhood rhabdomyosarcoma

BACKGROUND

Cross-sectional imaging-based morphological characteristics of pediatric rhabdomyosarcoma have failed to predict outcomes.

OBJECTIVE

To evaluate the feasibility and possible value of generating tumor sub-volumes using voxel-wise analysis of metabolic and functional data from positron emission tomography/magnetic resonance imaging (PET/MR) or PET/computed tomography (CT) and MRI in rhabdomyosarcoma.

MATERIALS AND METHODS

Thirty-four examinations in 17 patients who received PET/MRI or PET/CT plus MRI were analyzed. The volume of interest included total tumor volume before and after therapy. Apparent diffusion coefficients (ADC) and standard uptake values (SUV) were determined voxel-wise. Voxels were assigned to three different groups based on ADC and SUV: "viable tumor tissue," "intermediate tissue" or "possible necrosis." In a second approach, data were grouped into three clusters using the Gaussian mixture model. The ratio of these clusters to total tumor volume and changes due to chemotherapy were correlated with clinical and histopathological data.

RESULTS

After chemotherapy, the proportion of voxels in the different groups changed significantly. A significant reduction of the proportion of voxels assigned to cluster 1 was found, from a mean of 36.4% to 2.5% (P < 0.001). There was a significant increase in the proportion of voxels in cluster 3 following chemotherapy from 24.8% to 81.6% (P = 0.02). The proportion of voxels in cluster 2 differed depending on the presence or absence of tumor recurrence, falling from 48% to 10% post-chemotherapy in the group with no tumor recurrence (P < 0.05) and from 29% to 23% (P > 0.05) in the group with tumor recurrence.

CONCLUSION

Voxel-wise evaluation of multimodal data in rhabdomyosarcoma is feasible. Our initial results suggest that the different distribution of sub-volumes before and after therapy may have prognostic significance.

Span of regularization for solution of inverse problems with application to magnetic resonance relaxometry of the brain

We present a new regularization method for the solution of the Fredholm integral equation (FIE) of the first kind, in which we incorporate solutions corresponding to a range of Tikhonov regularizers into the end result. This method identifies solutions within a much larger function space, spanned by this set of regularized solutions, than is available to conventional regularization methods. An additional key development is the use of dictionary functions derived from noise-corrupted inversion of the discretized FIE. In effect, we combine the stability of solutions with greater degrees of regularization with the resolution of those that are less regularized. The span of regularizations (SpanReg) method may be widely applicable throughout the field of inverse problems.

Cryo-surgery for symptomatic extra-abdominal desmoids. A proof of concept study

BACKGROUND AND OBJECTIVE

Extra abdominal desmoid tumors are rare, highly aggressive, and invasive benign soft tissue tumors. Current treatment modalities show high levels of recurrence and comorbidities. Cryo-surgery as an alternative was subsequently investigated.

METHODS

In this retrospective, single center study 11 patients showing symptomatic tumors were treated with individualized cryo-surgery. Treatment protocol included preoperative planning using computer rendered 3D models, intraoperative navigation and execution using cone beam guidance, and postoperative magnetic resonance imaging image analysis using a gaussian mixture model software. Subjective outcomes were reported using Short Form Health Survey (SF-36) questionnaires.

RESULTS

Sixteen ablations were performed, each demonstrating a complete match with the determined preoperative plan and model. A total of 9/11 (82%) of patients showed improvements in symptoms and a reduction in tumor volume while 2/11 (18%) did not. Average reduction in tumor volume and viable segments were 36.7% (p = 0.0397) and 63.3% (p = 0.0477), respectively. Mild complications according to the SIR Adverse Event Classification Guidelines were experienced in 3/16 (19%) ablations. SF-36 scores showed a statistically significant improvement (p = 0.0194) in the mental health category and a nonsignificant (p = 0.8071) improvement in the physical health category.

CONCLUSION

Cryo-surgery using the three-phase protocol as described may improve the overall outcome of future ablation procedures.

Multiparametric MRI and Coregistered Histology Identify Tumor Habitats in Breast Cancer Mouse Models

It is well-recognized that solid tumors are genomically, anatomically, and physiologically heterogeneous. In general, more heterogeneous tumors have poorer outcomes, likely due to the increased probability of harboring therapy-resistant cells and regions. It is hypothesized that the genomic and physiologic heterogeneity are related, because physiologically distinct regions will exert variable selection pressures leading to the outgrowth of clones with variable genomic/proteomic profiles. To investigate this, methods must be in place to interrogate and define, at the microscopic scale, the cytotypes that exist within physiologically distinct subregions ("habitats") that are present at mesoscopic scales. MRI provides a noninvasive approach to interrogate physiologically distinct local environments, due to the biophysical principles that govern MRI signal generation. Here, we interrogate different physiologic parameters, such as perfusion, cell density, and edema, using multiparametric MRI (mpMRI). Signals from six different acquisition schema were combined voxel-by-voxel into four clusters identified using a Gaussian mixture model. These were compared with histologic and IHC characterizations of sections that were coregistered using MRI-guided 3D printed tumor molds. Specifically, we identified a specific set of MRI parameters to classify viable-normoxic, viable-hypoxic, nonviable-hypoxic, and nonviable-normoxic tissue types within orthotopic 4T1 and MDA-MB-231 breast tumors. This is the first coregistered study to show that mpMRI can be used to define physiologically distinct tumor habitats within breast tumor models. SIGNIFICANCE: This study demonstrates that noninvasive imaging metrics can be used to distinguish subregions within heterogeneous tumors with histopathologic correlation.

Data-driven identification of tumor subregions based on intravoxel incoherent motion reveals association with proliferative activity

PURPOSE

Intravoxel incoherent motion (IVIM) analysis gives information on tissue diffusion and perfusion and may thus have a potential for e.g. tumor tissue characterization. This work aims to study if clustering based on IVIM parameter maps can identify tumor subregions, and to assess the relevance of obtained subregions by histological analysis.

METHODS

Fourteen mice with human neuroendocrine tumors were examined with diffusion-weighted imaging to obtain IVIM parameter maps. Gaussian mixture models with IVIM maps from all tumors as input were used to partition voxels into k clusters, where k = 2 was chosen for further analysis based on goodness of fit. Clustering was performed with and without the perfusion-related IVIM parameter D * , and with and without including spatial information. The validity of the clustering was assessed by comparison with corresponding histologically stained tumor sections. A Ki-67-based index quantifying the degree of tumor proliferation was considered appropriate for the comparison based on the obtained cluster characteristics.

RESULTS

The clustering resulted in one class with low diffusion and high perfusion and another with slightly higher diffusion and low perfusion. Strong agreement was found between tumor subregions identified by clustering and subregions identified by histological analysis, both regarding size and spatial agreement. Neither D * nor spatial information had substantial effects on the clustering results.

CONCLUSIONS

The results of this study show that IVIM parameter maps can be used to identify tumor subregions using a data-driven framework based on Gaussian mixture models. In the studied tumor model, the obtained subregions showed agreement with proliferative activity.

DCE-MRI in Human Gliomas: A Surrogate for Assessment of Invasive Hypoxia Marker HIF-1Α Based on MRI-Neuronavigation Stereotactic Biopsies

RATIONALE AND OBJECTIVES

The purpose of this study was to correlate dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) parameters with data from a specific marker of hypoxia, hypoxia-inducible factor 1α (HIF-1α), in human gliomas on a point-to-point basis by using coregistered magnetic resonance imaging and frameless stereotactic biopsies.

MATERIALS AND METHODS

Thirty-four patients with treatment-naive gliomas underwent DCE, axial T1-weighted, T2-weighted, T2-weighted fluid acquisition of inversion recovery, and three-dimensional T1-weighted brain volume with gadolinium contrast enhancement sequences on a 3.0-T magnetic resonance scanner before stereotactic surgery. Quantitative perfusion indices such as endothelial transfer constant, fractional extravascular extracellular space volume, fractional plasma volume, and reflux rate were measured at corresponding stereotactic biopsy sites. Each sample was considered an independent measurement, and its histology grade was diagnosed. HIF-1α expression was quantified from the point-to-point biopsy tissues. Analyses of receiver operating characteristic curves were done for HIF-1α to discriminate different grades of glioma. To look for correlations between immunohistochemical parameters and DCE indices, Spearman's correlation coefficient was used.

RESULTS

Seventy biopsy samples from 34 subjects were included in the analysis. Mean immunoreactivity scores of HIF-1α were 2.75 ± 1.11 for grade II (n = 24), 6.20 ± 2.33 for grade III (n = 20), and 10.46 ± 2.42 for grade IV (n = 26). HIF-1α showed very good-to-excellent accuracy in discriminating grade II from III, III from IV, and II from IV (area under the curve = 0.838, 0.862, and 0.994, respectively). Endothelial transfer constant and fractional extravascular extracellular space volume showed a significantly positive correlation with HIF-1α expression (r = 0.686, P < .001; r = 0.549, P < .001, respectively).

CONCLUSION

Our study demonstrated HIF-1α to be a significant predictor of different grades of gliomas with high sensitivity and specificity. DCE-MRI is a useful, noninvasive imaging tool for quantitative evaluation of HIF-1α, and its parameters may be used as a surrogate for HIF-1α expression.

Ultrasound Shear Wave Elastography of the Normal Prostate: Interobserver Reproducibility and Comparison with Functional Magnetic Resonance Tissue Characteristics

The purpose of this study was to establish interobserver reproducibility of Young's modulus (YM) derived from ultrasound shear wave elastography (US-SWE) in the normal prostate and correlate it with multiparametric magnetic resonance imaging (mpMRI) tissue characteristics. Twenty men being screened for prostate cancer underwent same-day US-SWE (10 done by two blinded, newly-trained observers) and mpMRI followed by 12-core biopsy. Bland-Altman plots established limits of agreement for YM. Quantitative data from the peripheral zone (PZ) and the transitional zone (TZ) for YM, apparent diffusion coefficient (ADC, mm2/s from diffusion-weighted MRI), and Ktrans (volume transfer coefficient, min-1), Ve (extravascular-extracellular space, %), Kep (rate constant, /min), and initial area under the gadolinium concentration curve (IAUGC60, mmol/L/s) from dynamic contrast-enhanced MRI were obtained for slice-matched prostate sextants. Interobserver intraclass correlation coefficients were fair to good for individual regions (PZ = 0.57, TZ = 0.65) and for whole gland 0.67, (increasing to 0.81 when corrected for systematic observer bias). In the PZ, there were weak negative correlations between YM and ADC ( p = 0.008), and Ve ( p = 0.01) and a weak positive correlation with Kep ( p = 0.003). No significant intermodality correlations were seen in the TZ. Transrectal prostate US-SWE done without controlling manually applied probe pressure has fair/good interobserver reproducibility in inexperienced observers with potential to improve this to excellent by standardization of probe contact pressure. Within the PZ, increase in tissue stiffness is associated with reduced extracellular water (decreased ADC) and space (reduced Ve).

Uncertainty evaluation of image-based tumour control probability models in radiotherapy of prostate cancer using a visual analytic tool

Functional imaging techniques provide radiobiological information that can be included into tumour control probability (TCP) models to enable individualized outcome predictions in radiotherapy. However, functional imaging and the derived radiobiological information are influenced by uncertainties, translating into variations in individual TCP predictions. In this study we applied a previously developed analytical tool to quantify dose and TCP uncertainty bands when initial cell density is estimated from MRI-based apparent diffusion coefficient maps of eleven patients. TCP uncertainty bands of 16% were observed at patient level, while dose variations bands up to 8 Gy were found at voxel level for an iso-TCP approach.

Delineation of Tumor Habitats based on Dynamic Contrast Enhanced MRI

Tumor heterogeneity can be elucidated by mapping subregions of the lesion with differential imaging characteristics, called habitats. Dynamic Contrast Enhanced (DCE-)MRI can depict the tumor microenvironments by identifying areas with variable perfusion and vascular permeability, since individual tumor habitats vary in the rate and magnitude of the contrast uptake and washout. Of particular interest is identifying areas of hypoxia, characterized by inadequate perfusion and hyper-permeable vasculature. An automatic procedure for delineation of tumor habitats from DCE-MRI was developed as a two-part process involving: (1) statistical testing in order to determine the number of the underlying habitats; and (2) an unsupervised pattern recognition technique to recover the temporal contrast patterns and locations of the associated habitats. The technique is examined on simulated data and DCE-MRI, obtained from prostate and brain pre-clinical cancer models, as well as clinical data from sarcoma and prostate cancer patients. The procedure successfully identified habitats previously associated with well-perfused, hypoxic and/or necrotic tumor compartments. Given the association of tumor hypoxia with more aggressive tumor phenotypes, the obtained in vivo information could impact management of cancer patients considerably.

Multiparametric dynamic contrast-enhanced ultrasound imaging of prostate cancer

OBJECTIVES

The aim of this study is to improve the accuracy of dynamic contrast-enhanced ultrasound (DCE-US) for prostate cancer (PCa) localization by means of a multiparametric approach.

MATERIALS AND METHODS

Thirteen different parameters related to either perfusion or dispersion were extracted pixel-by-pixel from 45 DCE-US recordings in 19 patients referred for radical prostatectomy. Multiparametric maps were retrospectively produced using a Gaussian mixture model algorithm. These were subsequently evaluated on their pixel-wise performance in classifying 43 benign and 42 malignant histopathologically confirmed regions of interest, using a prostate-based leave-one-out procedure.

RESULTS

The combination of the spatiotemporal correlation (r), mean transit time (μ), curve skewness (κ), and peak time (PT) yielded an accuracy of 81% ± 11%, which was higher than the best performing single parameters: r (73%), μ (72%), and wash-in time (72%). The negative predictive value increased to 83% ± 16% from 70%, 69% and 67%, respectively. Pixel inclusion based on the confidence level boosted these measures to 90% with half of the pixels excluded, but without disregarding any prostate or region.

CONCLUSIONS

Our results suggest multiparametric DCE-US analysis might be a useful diagnostic tool for PCa, possibly supporting future targeting of biopsies or therapy. Application in other types of cancer can also be foreseen.

KEY POINTS

• DCE-US can be used to extract both perfusion and dispersion-related parameters. • Multiparametric DCE-US performs better in detecting PCa than single-parametric DCE-US. • Multiparametric DCE-US might become a useful tool for PCa localization.

Automation of pattern recognition analysis of dynamic contrast-enhanced MRI data to characterize intratumoral vascular heterogeneity

PURPOSE

To automate dynamic contrast-enhanced MRI (DCE-MRI) data analysis by unsupervised pattern recognition (PR) to enable spatial mapping of intratumoral vascular heterogeneity.

METHODS

Three steps were automated. First, the arrival time of the contrast agent at the tumor was determined, including a calculation of the precontrast signal. Second, four criteria-based algorithms for the slice-specific selection of number of patterns (NP) were validated using 109 tumor slices from subcutaneous flank tumors of five different tumor models. The criteria were: half area under the curve, standard deviation thresholding, percent signal enhancement, and signal-to-noise ratio (SNR). The performance of these criteria was assessed by comparing the calculated NP with the visually determined NP. Third, spatial assignment of single patterns and/or pattern mixtures was obtained by way of constrained nonnegative matrix factorization.

RESULTS

The determination of the contrast agent arrival time at the tumor slice was successfully automated. For the determination of NP, the SNR-based approach outperformed other selection criteria by agreeing >97% with visual assessment. The spatial localization of single patterns and pattern mixtures, the latter inferring tumor vascular heterogeneity at subpixel spatial resolution, was established successfully by automated assignment from DCE-MRI signal-versus-time curves.

CONCLUSION

The PR-based DCE-MRI analysis was successfully automated to spatially map intratumoral vascular heterogeneity. Magn Reson Med 79:1736-1744, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

Decoding Intratumoral Heterogeneity of Breast Cancer by Multiparametric In Vivo Imaging: A Translational Study

Differential diagnosis and therapy of heterogeneous breast tumors poses a major clinical challenge. To address the need for a comprehensive, noninvasive strategy to define the molecular and functional profiles of tumors in vivo, we investigated a novel combination of metabolic PET and diffusion-weighted (DW)-MRI in the polyoma virus middle T antigen transgenic mouse model of breast cancer. The implementation of a voxelwise analysis for the clustering of intra- and intertumoral heterogeneity in this model resulted in a multiparametric profile based on [(18)F]Fluorodeoxyglucose ([(18)F]FDG)-PET and DW-MRI, which identified three distinct tumor phenotypes in vivo, including solid acinar, and solid nodular malignancies as well as cystic hyperplasia. To evaluate the feasibility of this approach for clinical use, we examined estrogen receptor-positive and progesterone receptor-positive breast tumors from five patient cases using DW-MRI and [(18)F]FDG-PET in a simultaneous PET/MRI system. The postsurgical in vivo PET/MRI data were correlated to whole-slide histology using the latter traditional diagnostic standard to define phenotype. By this approach, we showed how molecular, structural (microscopic, anatomic), and functional information could be simultaneously obtained noninvasively to identify precancerous and malignant subtypes within heterogeneous tumors. Combined with an automatized analysis, our results suggest that multiparametric molecular and functional imaging may be capable of providing comprehensive tumor profiling for noninvasive cancer diagnostics. Cancer Res; 76(18); 5512-22. ©2016 AACR.

A comparison of individual and population-derived vascular input functions for quantitative DCE-MRI in rats

Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) can quantitatively and qualitatively assess physiological characteristics of tissue. Quantitative DCE-MRI requires an estimate of the time rate of change of the concentration of the contrast agent in the blood plasma, the vascular input function (VIF). Measuring the VIF in small animals is notoriously difficult as it requires high temporal resolution images limiting the achievable number of slices, field-of-view, spatial resolution, and signal-to-noise. Alternatively, a population-averaged VIF could be used to mitigate the acquisition demands in studies aimed to investigate, for example, tumor vascular characteristics. Thus, the overall goal of this manuscript is to determine how the kinetic parameters estimated by a population based VIF differ from those estimated by an individual VIF. Eight rats bearing gliomas were imaged before, during, and after an injection of Gd-DTPA. K(trans), ve, and vp were extracted from signal-time curves of tumor tissue using both individual and population-averaged VIFs. Extended model voxel estimates of K(trans) and ve in all animals had concordance correlation coefficients (CCC) ranging from 0.69 to 0.98 and Pearson correlation coefficients (PCC) ranging from 0.70 to 0.99. Additionally, standard model estimates resulted in CCCs ranging from 0.81 to 0.99 and PCCs ranging from 0.98 to 1.00, supporting the use of a population based VIF if an individual VIF is not available.