Evangelia I. Zacharaki

Classification of brain tumor type and grade using MRI texture and shape in a machine learning scheme

The objective of this study is to investigate the use of pattern classification methods for distinguishing different types of brain tumors, such as primary gliomas from metastases, and also for grading of gliomas. The availability of an automated computer analysis tool that is more objective than human readers can potentially lead to more reliable and reproducible brain tumor diagnostic procedures. A computer‐assisted classification method combining conventional MRI and perfusion MRI is developed and used for differential diagnosis. The proposed scheme consists of several steps including region‐of‐interest definition, feature extraction, feature selection, and classification. The extracted features include tumor shape and intensity characteristics, as well as rotation invariant texture features. Feature subset selection is performed using support vector machines with recursive feature elimination. The method was applied on a population of 102 brain tumors histologically diagnosed as metastasis ( 24 ), meningiomas ( 4 ), gliomas World Health Organization grade II ( 22 ), gliomas World Health Organization grade III ( 18 ), and glioblastomas ( 34 ). The binary support vector machine classification accuracy, sensitivity, and specificity, assessed by leave‐one‐out cross‐validation, were, respectively, 85%, 87%, and 79% for discrimination of metastases from gliomas and 88%, 85%, and 96% for discrimination of high‐grade (grades III and IV) from low‐grade (grade II) neoplasms. Multiclass classification was also performed via a one‐vs‐all voting scheme. Magn Reson Med, 2009.

Multiparametric Tissue Characterization of Brain Neoplasms and Their Recurrence Using Pattern Classification of MR Images

RATIONALE AND OBJECTIVES Treatment of brain neoplasms can greatly benefit from better delineation of bulk neoplasm boundary and the extent and degree of more subtle neoplastic infiltration. Magnetic resonance imaging (MRI) is the primary imaging modality for evaluation before and after therapy, typically combining conventional sequences with more advanced techniques such as perfusion-weighted imaging and diffusion tensor imaging (DTI). The purpose of this study is to quantify the multiparametric imaging profile of neoplasms by integrating structural MRI and DTI via statistical image analysis methods to potentially capture complex and subtle tissue characteristics that are not obvious from any individual image or parameter. MATERIALS AND METHODS Five structural MRI sequences, namely, B0, diffusion-weighted images, fluid-attenuated inversion recovery, T1-weighted, and gadolinium-enhanced T1-weighted, and two scalar maps computed from DTI (ie, fractional anisotropy and apparent diffusion coefficient) are used to create an intensity-based tissue profile. This is incorporated into a nonlinear pattern classification technique to create a multiparametric probabilistic tissue characterization, which is applied to data from 14 patients with newly diagnosed primary high-grade neoplasms who have not received any therapy before imaging. RESULTS Preliminary results demonstrate that this multiparametric tissue characterization helps to better differentiate among neoplasm, edema, and healthy tissue, and to identify tissue that is likely to progress to neoplasm in the future. This has been validated on expert assessed tissue. CONCLUSION This approach has potential applications in treatment, aiding computer-assisted surgery by determining the spatial distributions of healthy and neoplastic tissue, as well as in identifying tissue that is relatively more prone to tumor recurrence.

Deformable registration of brain tumor images via a statistical model of tumor-induced deformation

An approach to deformable registration of three-dimensional brain tumor images to a normal brain atlas is presented. The approach involves the integration of three components: a biomechanical model of tumor mass-effect, a statistical approach to estimate the models parameters, and a deformable image registration method. Statistical properties of the desired deformation map are first obtained through tumor mass-effect simulations on normal brain images. This map is decomposed into the sum of two components in orthogonal subspaces, one representing inter-individual differences, and the other involving tumor-induced deformation. For a new tumor case, a partial observation of the desired deformation map is obtained via deformable image registration and is decomposed into the aforementioned spaces in order to estimate the mass-effect model parameters. Using this estimate, a simulation of tumor mass-effect is performed on the atlas to generate an image that is more similar to brain tumor image, thereby facilitating the atlas registration process. Results for a real and a simulated tumor case indicate significant reduction in the registration error due to the presented approach as compared to the direct use of deformable image registration.

ORBIT: A Multiresolution Framework for Deformable Registration of Brain Tumor Images

A deformable registration method is proposed for registering a normal brain atlas with images of brain tumor patients. The registration is facilitated by first simulating the tumor mass effect in the normal atlas in order to create an atlas image that is as similar as possible to the patients image. An optimization framework is used to optimize the location of tumor seed as well as other parameters of the tumor growth model, based on the pattern of deformation around the tumor region. In particular, the optimization is implemented in a multiresolution and hierarchical scheme, and it is accelerated by using a principal component analysis (PCA)-based model of tumor growth and mass effect, trained on a computationally more expensive biomechanical model. Validation on simulated and real images shows that the proposed registration framework, referred to as ORBIT (optimization of tumor parameters and registration of brain images with tumors), outperforms other available registration methods particularly for the regions close to the tumor, and it has the potential to assist in constructing statistical atlases from tumor-diseased brain images.

Non-diffeomorphic registration of brain tumor images by simulating tissue loss and tumor growth.

Although a variety of diffeomorphic deformable registration methods exist in the literature, application of these methods in the presence of space-occupying lesions is not straightforward. The motivation of this work is spatial normalization of MR images from patients with brain tumors in a common stereotaxic space, aiming to pool data from different patients into a common space in order to perform group analyses. Additionally, transfer of structural and functional information from neuroanatomical brain atlases into the individual patients space can be achieved via the inverse mapping, for the purpose of segmenting brains and facilitating surgical or radiotherapy treatment planning. A method that estimates the brain tissue loss and replacement by tumor is applied for achieving equivalent image content between an atlas and a patients scan, based on a biomechanical model of tumor growth. Automated estimation of the parameters modeling brain tissue loss and displacement is performed via optimization of an objective function reflecting feature-based similarity and elastic stretching energy, which is optimized in parallel via APPSPACK (Asynchronous Parallel Pattern Search). The results of the method, applied to 21 brain tumor patients, indicate that the registration accuracy is relatively high in areas around the tumor, as well as in the healthy portion of the brain. Also, the calculated deformation in the vicinity of the tumor is shown to correlate highly with expert-defined visual scores indicating the tumor mass effect, thereby potentially leading to an objective approach to quantification of mass effect, which is commonly used in diagnosis.

In silico radiation oncology: combining novel simulation algorithms with current visualization techniques

The concept of in silica radiation oncology is clarified in this paper. A brief literature review points out the principal domains in which experimental, mathematical, and three-dimensional (3-D) computer simulation models of tumor growth and response to radiation therapy have been developed. Two paradigms of 3-D simulation models developed by our research group are concisely presented. The first one refers to the in vitro development and radiation response of a tumor spheroid whereas the second one refers to the fractionated radiation response of a clinical tumor in vivo based on the patients imaging data. In each case, a description of the salient points of the corresponding algorithms and the visualization techniques used takes place. Specific applications of the models to experimental and clinical cases are described and the behavior of the models is two- and three-dimensionally visualized by using virtual reality techniques. Good qualitative agreement with experimental and clinical observations strengthens the applicability of the models to real situations. A protocol for further testing and adaptation is outlined. Therefore, an advanced integrated patient specific decision support and spatio-temporal treatment planning system is expected to emerge after the completion of the necessary experimental tests and clinical evaluation.

STEP: Spatiotemporal enhancement pattern for MR-based breast tumor diagnosis

The authors propose a spatiotemporal enhancement pattern (STEP) for comprehensive characterization of breast tumors in contrast-enhanced MR images. By viewing serial contrast-enhanced MR images as a single spatiotemporal image, they formulate the STEP as a combination of (1) dynamic enhancement and architectural features of a tumor, and (2) the spatial variations of pixelwise temporal enhancements. Although the latter has been widely used by radiologists for diagnostic purposes, it has rarely been employed for computer-aided diagnosis. This article presents two major contributions. First, the STEP features are introduced to capture temporal enhancement and its spatial variations. This is essentially carried out through the Fourier transformation and pharmacokinetic modeling of various temporal enhancement features, followed by the calculation of moment invariants and Gabor texture features. Second, for effectively extracting the STEP features from tumors, we develop a graph-cut based segmentation algorithm that aims at refining coarse manual segmentations of tumors. The STEP features are assessed through their diagnostic performance for differentiating between benign and malignant tumors using a linear classifier (along with a simple ranking-based feature selection) in a leave-one-out cross-validation setting. The experimental results for the proposed features exhibit superior performance, when compared to the existing approaches, with the area under the ROC curve approaching 0.97.

Survival Analysis of Patients with High-Grade Gliomas Based on Data Mining of Imaging Variables

The authors compared survival prediction using machine-learning data-mining techniques versus traditional histopathologic analysis in a group of 74 tumors. They looked at 55 imaging variables, extracting the most important ones and comparing these with standard histology. The most important variables were: extent of resection, mass effects, volume of enhancing tumor, maximum T2 intensity, and mean trace intensity in non-enhancing or edematous surrounding areas. The accuracy of the data-mining method was 85% and based on Kaplan-Meier curves this predicted survival better than histology. BACKGROUND AND PURPOSE: The prediction of prognosis in HGGs is poor in the majority of patients. Our aim was to test whether multivariate prediction models constructed by machine-learning methods provide a more accurate predictor of prognosis in HGGs than histopathologic classification. The prediction of survival was based on DTI and rCBV measurements as an adjunct to conventional imaging. MATERIALS AND METHODS: The relationship of survival to 55 variables, including clinical parameters (age, sex), categoric or continuous tumor descriptors (eg, tumor location, extent of resection, multifocality, edema), and imaging characteristics in ROIs, was analyzed in a multivariate fashion by using data-mining techniques. A variable selection method was applied to identify the overall most important variables. The analysis was performed on 74 HGGs (18 anaplastic gliomas WHO grades III/IV and 56 GBMs or gliosarcomas WHO grades IV/IV). RESULTS: Five variables were identified as the most significant, including the extent of resection, mass effect, volume of enhancing tumor, maximum B0 intensity, and mean trace intensity in the nonenhancing/edematous region. These variables were used to construct a prediction model based on a J48 classification tree. The average classification accuracy, assessed by cross-validation, was 85.1%. Kaplan-Meier survival curves showed that the constructed prediction model classified malignant gliomas in a manner that better correlates with clinical outcome than standard histopathology. CONCLUSIONS: Prediction models based on data-mining algorithms can provide a more accurate predictor of prognosis in malignant gliomas than histopathologic classification alone.

Measuring Brain Lesion Progression with a Supervised Tissue Classification System

Brain lesions, especially White Matter Lesions (WMLs), are associated with cardiac and vascular disease, but also with normal aging. Quantitative analysis of WML in large clinical trials is becoming more and more important. In this paper, we present a computer-assisted WML segmentation method, based on local features extracted from conventional multi-parametric Magnetic Resonance Imaging (MRI) sequences. A framework for preprocessing the temporal data by jointly equalizing histograms reduces the spatial and temporal variance of data, thereby improving the longitudinal stability of such measurements and hence the estimate of lesion progression. A Support Vector Machine (SVM) classifier trained on expert-defined WMLs is applied for lesion segmentation on each scan using the AdaBoost algorithm. Validation on a population of 23 patients from 3 different imaging sites with follow-up studies and WMLs of varying sizes, shapes and locations tests the robustness and accuracy of the proposed segmentation method, compared to the manual segmentation results from an experienced neuroradiologist. The results show that our CAD-system achieves consistent lesion segmentation in the 4D data facilitating the disease monitoring.

Promises and challenges for the implementation of computational medical imaging (radiomics) in oncology

Medical image processing and analysis (also known as Radiomics) is a rapidly growing discipline that maps digital medical images into quantitative data, with the end goal of generating imaging biomarkers as decision support tools for clinical practice. The use of imaging data from routine clinical work-up has tremendous potential in improving cancer care by heightening understanding of tumor biology and aiding in the implementation of precision medicine. As a noninvasive method of assessing the tumor and its microenvironment in their entirety, radiomics allows the evaluation and monitoring of tumor characteristics such as temporal and spatial heterogeneity. One can observe a rapid increase in the number of computational medical imaging publications-milestones that have highlighted the utility of imaging biomarkers in oncology. Nevertheless, the use of radiomics as clinical biomarkers still necessitates amelioration and standardization in order to achieve routine clinical adoption. This Review addresses the critical issues to ensure the proper development of radiomics as a biomarker and facilitate its implementation in clinical practice.