Quantification of intratumoural heterogeneity in mice and patients via machine-learning models trained on PET-MRI data.

In oncology, intratumoural heterogeneity is closely linked with the efficacy of therapy, and can be partially characterized via tumour biopsies. Here we show that intratumoural heterogeneity can be characterized spatially via phenotype-specific, multi-view learning classifiers trained with data from dynamic positron emission tomography (PET) and multiparametric magnetic resonance imaging (MRI). Classifiers trained with PET-MRI data from mice with subcutaneous colon cancer quantified phenotypic changes resulting from an apoptosis-inducing targeted therapeutic and provided biologically relevant probability maps of tumour-tissue subtypes. When applied to retrospective PET-MRI data of patients with liver metastases from colorectal cancer, the trained classifiers characterized intratumoural tissue subregions in agreement with tumour histology. The spatial characterization of intratumoural heterogeneity in mice and patients via multimodal, multiparametric imaging aided by machine-learning may facilitate applications in precision oncology.

Machine learning identifies stroke features between species.

Identification and localization of ischemic stroke (IS) lesions is routinely performed to confirm diagnosis, assess stroke severity, predict disability and plan rehabilitation strategies using magnetic resonance imaging (MRI). In basic research, stroke lesion segmentation is necessary to study complex peri-infarction tissue changes. Moreover, final stroke volume is a critical outcome evaluated in clinical and preclinical experiments to determine therapy or intervention success. Manual segmentations are performed but they require a specialized skill set, are prone to inter-observer variation, are not entirely objective and are often not supported by histology. The task is even more challenging when dealing with large multi-center datasets, multiple experimenters or large animal cohorts. On the other hand, current automatized segmentation approaches often lack histological validation, are not entirely user independent, are often based on single parameters, or in the case of complex machine learning methods, require vast training datasets and are prone to a lack of model interpretation. Methods: We induced IS using the middle cerebral artery occlusion model on two rat cohorts. We acquired apparent diffusion coefficient (ADC) and T2-weighted (T2W) images at 24 h and 1-week after IS induction. Subsets of the animals at 24 h and 1-week post IS were evaluated using histology and immunohistochemistry. Using a Gaussian mixture model, we segmented voxel-wise interactions between ADC and T2W parameters at 24 h using one of the rat cohorts. We then used these segmentation results to train a random forest classifier, which we applied to the second rat cohort. The algorithms' stroke segmentations were compared to manual stroke delineations, T2W and ADC thresholding methods and the final stroke segmentation at 1-week. Volume correlations to histology were also performed for every segmentation method. Metrics of success were calculated with respect to the final stroke volume. Finally, the trained random forest classifier was tested on a human dataset with a similar temporal stroke on-set. Manual segmentations, ADC and T2W thresholds were again used to evaluate and perform comparisons with the proposed algorithms' output. Results: In preclinical rat data our framework significantly outperformed commonly applied automatized thresholding approaches and segmented stroke regions similarly to manual delineation. The framework predicted the localization of final stroke regions in 1-week post-stroke MRI with a median Dice similarity coefficient of 0.86, Matthew's correlation coefficient of 0.80 and false positive rate of 0.04. The predicted stroke volumes also strongly correlated with final histological stroke regions (Pearson correlation = 0.88, P < 0.0001). Lastly, the stroke region characteristics identified by our framework in rats also identified stroke lesions in human brains, largely outperforming thresholding approaches in stroke volume prediction (P<0.01). Conclusion: Our findings reveal that the segmentation produced by our proposed framework using 24 h MRI rat data strongly correlated with the final stroke volume, denoting a predictive effect. In addition, we show for the first time that the stroke imaging features can be directly translated between species, allowing identification of acute stroke in humans using the model trained on animal data. This discovery reduces the gap between the clinical and preclinical fields, unveiling a novel approach to directly co-analyze clinical and preclinical data. Such methods can provide further biological insights into human stroke and highlight the differences between species in order to help improve the experimental setups and animal models of the disease.

PET/MRI Hybrid Systems.

Over the last decade, the combination of PET and MRI in one system has proven to be highly successful in basic preclinical research, as well as in clinical research. Nowadays, PET/MRI systems are well established in preclinical imaging and are progressing into clinical applications to provide further insights into specific diseases, therapeutic assessments, and biological pathways. Certain challenges in terms of hardware had to be resolved concurrently with the development of new techniques to be able to reach the full potential of both combined techniques. This review provides an overview of these challenges and describes the opportunities that simultaneous PET/MRI systems can exploit in comparison with stand-alone or other combined hybrid systems. New approaches were developed for simultaneous PET/MRI systems to correct for attenuation of 511 keV photons because MRI does not provide direct information on gamma photon attenuation properties. Furthermore, new algorithms to correct for motion were developed, because MRI can accurately detect motion with high temporal resolution. The additional information gained by the MRI can be employed to correct for partial volume effects as well. The development of new detector designs in combination with fast-decaying scintillator crystal materials enabled time-of-flight detection and incorporation in the reconstruction algorithms. Furthermore, this review lists the currently commercially available systems both for preclinical and clinical imaging and provides an overview of applications in both fields. In this regard, special emphasis has been placed on data analysis and the potential for both modalities to evolve with advanced image analysis tools, such as cluster analysis and machine learning.

Impact of the Arterial Input Function Recording Method on Kinetic Parameters in Small-Animal PET.

The goal of this study was to validate the use of an MR-compatible blood sampler (BS) with a detector system based on a lutetium oxyorthosilicate scintillator and avalanche photodiodes for small-animal PET. Methods: Five rats underwent a 60-min 18F-FDG study. For each animal, the arterial input function (AIF) was derived from the BS recording, from manual sampling (MS), and from the PET image. These AIFs were applied for kinetic modeling of the striatum using the irreversible 2-tissue-compartment model. The MS-based technique with a dispersion correction served as a reference approach, and the kinetic parameters that were estimated with the BS- and the image-derived AIFs were compared with the reference values. Additionally, the effect of applying a population-based activity ratio for plasma to whole blood (p/wb) and the dispersion correction was assessed. Results: The K1, k2, and k3 values estimated with the reference approach were 0.174 ± 0.037 mL/min/cm3, 0.342 ± 0.080 1/min, and 0.048 ± 0.009 1/min, respectively. The corresponding parameters obtained with the BS- and image-derived AIFs deviated from these values by 0.6%-18.8% and 16.7%-47.9%, respectively. To compensate for the error in the BS-based technique, data from one MS collected at the end of the experiment were combined with the data from the first 10 min of the BS recording. This approach reduced the deviation in the kinetic parameters to 1.8%-6.3%. Using p/wb led to a 1.7%-8.3% difference from the reference parameters. The sensitivity of the BS was 23%, the energy resolution for the 511-keV photopeak was 19%, and the timing resolution was 11.2 ns. Conclusion: Online recording of the blood activity level with the BS allows precise measurement of AIF, without loss of blood volume. Combining the BS data with one MS is the most accurate approach for the data analysis. The high sensitivity of the device may allow application of lower radioactivity doses.

A Novel Unsupervised Segmentation Approach Quantifies Tumor Tissue Populations Using Multiparametric MRI: First Results with Histological Validation.

PURPOSE: We aimed to precisely estimate intra-tumoral heterogeneity using spatially regularized spectral clustering (SRSC) on multiparametric MRI data and compare the efficacy of SRSC with the previously reported segmentation techniques in MRI studies. PROCEDURES: Six NMRI nu/nu mice bearing subcutaneous human glioblastoma U87 MG tumors were scanned using a dedicated small animal 7T magnetic resonance imaging (MRI) scanner. The data consisted of T2 weighted images, apparent diffusion coefficient maps, and pre- and post-contrast T2 and T2* maps. Following each scan, the tumors were excised into 2-3-mm thin slices parallel to the axial field of view and processed for histological staining. The MRI data were segmented using SRSC, K-means, fuzzy C-means, and Gaussian mixture modeling to estimate the fractional population of necrotic, peri-necrotic, and viable regions and validated with the fractional population obtained from histology. RESULTS: While the aforementioned methods overestimated peri-necrotic and underestimated viable fractions, SRSC accurately predicted the fractional population of all three tumor tissue types and exhibited strong correlations (rnecrotic = 0.92, rperi-necrotic = 0.82 and rviable = 0.98) with the histology. CONCLUSIONS: The precise identification of necrotic, peri-necrotic and viable areas using SRSC may greatly assist in cancer treatment planning and add a new dimension to MRI-guided tumor biopsy procedures.

Spectral Clustering Predicts Tumor Tissue Heterogeneity Using Dynamic 18F-FDG PET: A Complement to the Standard Compartmental Modeling Approach.

In this study, we described and validated an unsupervised segmentation algorithm for the assessment of tumor heterogeneity using dynamic 18F-FDG PET. The aim of our study was to objectively evaluate the proposed method and make comparisons with compartmental modeling parametric maps and SUV segmentations using simulations of clinically relevant tumor tissue types. Methods: An irreversible 2-tissue-compartmental model was implemented to simulate clinical and preclinical 18F-FDG PET time-activity curves using population-based arterial input functions (80 clinical and 12 preclinical) and the kinetic parameter values of 3 tumor tissue types. The simulated time-activity curves were corrupted with different levels of noise and used to calculate the tissue-type misclassification errors of spectral clustering (SC), parametric maps, and SUV segmentation. The utility of the inverse noise variance- and Laplacian score-derived frame weighting schemes before SC was also investigated. Finally, the SC scheme with the best results was tested on a dynamic 18F-FDG measurement of a mouse bearing subcutaneous colon cancer and validated using histology. Results: In the preclinical setup, the inverse noise variance-weighted SC exhibited the lowest misclassification errors (8.09%-28.53%) at all noise levels in contrast to the Laplacian score-weighted SC (16.12%-31.23%), unweighted SC (25.73%-40.03%), parametric maps (28.02%-61.45%), and SUV (45.49%-45.63%) segmentation. The classification efficacy of both weighted SC schemes in the clinical case was comparable to the unweighted SC. When applied to the dynamic 18F-FDG measurement of colon cancer, the proposed algorithm accurately identified densely vascularized regions from the rest of the tumor. In addition, the segmented regions and clusterwise average time-activity curves showed excellent correlation with the tumor histology. Conclusion: The promising results of SC mark its position as a robust tool for quantification of tumor heterogeneity using dynamic PET studies. Because SC tumor segmentation is based on the intrinsic structure of the underlying data, it can be easily applied to other cancer types as well.

A Population-Based Gaussian Mixture Model Incorporating 18F-FDG PET and Diffusion-Weighted MRI Quantifies Tumor Tissue Classes.

UNLABELLED: The aim of our study was to create a novel Gaussian mixture modeling (GMM) pipeline to model the complementary information derived from(18)F-FDG PET and diffusion-weighted MRI (DW-MRI) to separate the tumor microenvironment into relevant tissue compartments and follow the development of these compartments longitudinally. METHODS: Serial (18)F-FDG PET and apparent diffusion coefficient (ADC) maps derived from DW-MR images of NCI-H460 xenograft tumors were coregistered, and a population-based GMM was implemented on the complementary imaging data. The tumor microenvironment was segmented into 3 distinct regions and correlated with histology. ANCOVA was applied to gauge how well the total tumor volume was a predictor for the ADC and (18)F-FDG, or if ADC was a good predictor of (18)F-FDG for average values in the whole tumor or average necrotic and viable tissues. RESULTS: The coregistered PET/MR images were in excellent agreement with histology, both visually and quantitatively, and allowed for validation of the last-time-point measurements. Strong correlations were found for the necrotic (r = 0.88) and viable fractions (r = 0.87) between histology and clustering. The GMM provided probabilities for each compartment with uncertainties expressed as a mixture of tissues in which the resolution of scans was inadequate to accurately separate tissues. The ANCOVA suggested that both ADC and (18)F-FDG in the whole tumor (P = 0.0009, P = 0.02) as well as necrotic (P = 0.008, P = 0.02) and viable (P = 0.003, P = 0.01) tissues were a positive, linear function of total tumor volume. ADC proved to be a positive predictor of (18)F-FDG in the whole tumor (P = 0.001) and necrotic (P = 0.02) and viable (P = 0.0001) tissues. CONCLUSION: The complementary information of (18)F-FDG and ADC longitudinal measurements in xenograft tumors allows for segmentation into distinct tissues when using the novel GMM pipeline. Leveraging the power of multiparametric PET/MRI in this manner has the potential to take the assessment of disease outcome beyond RECIST and could provide an important impact to the field of precision medicine.