skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: SU-F-R-17: Advancing Glioblastoma Multiforme (GBM) Recurrence Detection with MRI Image Texture Feature Extraction and Machine Learning

Abstract

Purpose: To test the potential of early Glioblastoma Multiforme (GBM) recurrence detection utilizing image texture pattern analysis in serial MR images post primary treatment intervention. Methods: MR image-sets of six time points prior to the confirmed recurrence diagnosis of a GBM patient were included in this study, with each time point containing T1 pre-contrast, T1 post-contrast, T2-Flair, and T2-TSE images. Eight Gray-level co-occurrence matrix (GLCM) texture features including Contrast, Correlation, Dissimilarity, Energy, Entropy, Homogeneity, Sum-Average, and Variance were calculated from all images, resulting in a total of 32 features at each time point. A confirmed recurrent volume was contoured, along with an adjacent non-recurrent region-of-interest (ROI) and both volumes were propagated to all prior time points via deformable image registration. A support vector machine (SVM) with radial-basis-function kernels was trained on the latest time point prior to the confirmed recurrence to construct a model for recurrence classification. The SVM model was then applied to all prior time points and the volumes classified as recurrence were obtained. Results: An increase in classified volume was observed over time as expected. The size of classified recurrence maintained at a stable level of approximately 0.1 cm{sup 3} up to 272 days prior to confirmation.more » Noticeable volume increase to 0.44 cm{sup 3} was demonstrated at 96 days prior, followed by significant increase to 1.57 cm{sup 3} at 42 days prior. Visualization of the classified volume shows the merging of recurrence-susceptible region as the volume change became noticeable. Conclusion: Image texture pattern analysis in serial MR images appears to be sensitive to detecting the recurrent GBM a long time before the recurrence is confirmed by a radiologist. The early detection may improve the efficacy of targeted intervention including radiosurgery. More patient cases will be included to create a generalizable classification model applicable to a larger patient cohort. NIH R43CA183390 and R01CA188300.NSF Graduate Research Fellowship DGE-1144087.« less

Authors:
; ; ; ; ;  [1]
  1. UCLA School of Medicine, Los Angeles, CA (United States)
Publication Date:
OSTI Identifier:
22626742
Resource Type:
Journal Article
Resource Relation:
Journal Name: Medical Physics; Journal Volume: 43; Journal Issue: 6; Other Information: (c) 2016 American Association of Physicists in Medicine; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
60 APPLIED LIFE SCIENCES; 61 RADIATION PROTECTION AND DOSIMETRY; ENTROPY; GLIOMAS; IMAGES; NMR IMAGING; PATIENTS; RADIOTHERAPY; SURGERY

Citation Formats

Yu, V, Ruan, D, Nguyen, D, Kaprealian, T, Chin, R, and Sheng, K. SU-F-R-17: Advancing Glioblastoma Multiforme (GBM) Recurrence Detection with MRI Image Texture Feature Extraction and Machine Learning. United States: N. p., 2016. Web. doi:10.1118/1.4955789.
Yu, V, Ruan, D, Nguyen, D, Kaprealian, T, Chin, R, & Sheng, K. SU-F-R-17: Advancing Glioblastoma Multiforme (GBM) Recurrence Detection with MRI Image Texture Feature Extraction and Machine Learning. United States. doi:10.1118/1.4955789.
Yu, V, Ruan, D, Nguyen, D, Kaprealian, T, Chin, R, and Sheng, K. 2016. "SU-F-R-17: Advancing Glioblastoma Multiforme (GBM) Recurrence Detection with MRI Image Texture Feature Extraction and Machine Learning". United States. doi:10.1118/1.4955789.
@article{osti_22626742,
title = {SU-F-R-17: Advancing Glioblastoma Multiforme (GBM) Recurrence Detection with MRI Image Texture Feature Extraction and Machine Learning},
author = {Yu, V and Ruan, D and Nguyen, D and Kaprealian, T and Chin, R and Sheng, K},
abstractNote = {Purpose: To test the potential of early Glioblastoma Multiforme (GBM) recurrence detection utilizing image texture pattern analysis in serial MR images post primary treatment intervention. Methods: MR image-sets of six time points prior to the confirmed recurrence diagnosis of a GBM patient were included in this study, with each time point containing T1 pre-contrast, T1 post-contrast, T2-Flair, and T2-TSE images. Eight Gray-level co-occurrence matrix (GLCM) texture features including Contrast, Correlation, Dissimilarity, Energy, Entropy, Homogeneity, Sum-Average, and Variance were calculated from all images, resulting in a total of 32 features at each time point. A confirmed recurrent volume was contoured, along with an adjacent non-recurrent region-of-interest (ROI) and both volumes were propagated to all prior time points via deformable image registration. A support vector machine (SVM) with radial-basis-function kernels was trained on the latest time point prior to the confirmed recurrence to construct a model for recurrence classification. The SVM model was then applied to all prior time points and the volumes classified as recurrence were obtained. Results: An increase in classified volume was observed over time as expected. The size of classified recurrence maintained at a stable level of approximately 0.1 cm{sup 3} up to 272 days prior to confirmation. Noticeable volume increase to 0.44 cm{sup 3} was demonstrated at 96 days prior, followed by significant increase to 1.57 cm{sup 3} at 42 days prior. Visualization of the classified volume shows the merging of recurrence-susceptible region as the volume change became noticeable. Conclusion: Image texture pattern analysis in serial MR images appears to be sensitive to detecting the recurrent GBM a long time before the recurrence is confirmed by a radiologist. The early detection may improve the efficacy of targeted intervention including radiosurgery. More patient cases will be included to create a generalizable classification model applicable to a larger patient cohort. NIH R43CA183390 and R01CA188300.NSF Graduate Research Fellowship DGE-1144087.},
doi = {10.1118/1.4955789},
journal = {Medical Physics},
number = 6,
volume = 43,
place = {United States},
year = 2016,
month = 6
}
  • Purpose: To determine whether the combined magnetic resonance imaging (MRI) and magnetic resonance spectroscopy imaging (MRSI) before radiation therapy (RT) is valuable for RT target definition, and to evaluate the feasibility of replacing the current definition of uniform margins by custom-shaped margins based on the information from MRI and MRSI. Methods and Materials: A total of 23 glioblastoma multiforme (GBM) patients underwent MRI and MRSI within 4 weeks after surgery but before the initiation of RT and at 2-month follow-up intervals thereafter. The MRSI data were quantified on the basis of a Choline-to-NAA Index (CNI) as a measure of spectroscopicmore » abnormality. A combined anatomic and metabolic region of interest (MRI/S) consisting of T2-weighted hyperintensity, contrast enhancement (CE), resection cavity, and CNI2 (CNI {>=} 2) based on the pre-RT imaging was compared to the extent of CNI2 and the RT dose distribution. The spatial relationship of the pre-RT MRI/S and the RT dose volume was compared with the extent of CE at each follow-up. Results: Nine patients showed new or increased CE during follow-up, and 14 patients were either stable or had decreased CE. New or increased areas of CE occurred within CNI2 that was covered by 60 Gy in 6 patients and within the CNI2 that was not entirely covered by 60 Gy in 3 patients. New or increased CE resided within the pre-RT MRI/S lesion in 89% (8/9) of the patients with new or increased CE. Conclusion: These data indicate that the definition of RT target volumes according to the combined morphologic and metabolic abnormality may be sufficient for RT targeting.« less
  • The aim of the present study was to detect complex genetic alterations in human glioblastoma multiforme (GBM) by comparative genomic in situ hybridization (CGH). Of the 24 GBM that were examined, increased fluorescence intensities indicating chromosomal polysomy of chromosome 7 and gene amplification at chromosome 7p were found in 42% of the tumors. In addition, signal enhancement of chromosome 19 was present in 29% and at 12q13-15 in 21% of the tumors. We also detected reduction of fluorescence intensities indicating gross deletions on chromosomes 10 (58%), 9p (46%), and 13 (29%). There was a close correlation of CGH results whenmore » compared with Southern analysis of the EGFR gene localized on chromosome 7 and loss of heterozygosity detection of chromosome 9 and 10 by microsatellite PCR. A close correlation was also observed between copy number changes of chromosome 7 and deletions of chromosome 10. Amplification of chromosome 12q and deletions of chromosomes 9p and 13 seemed to be complementary in the tumors investigated in the present study. 44 refs., 3 figs., 1 tab.« less
  • ‘Radiomics’ refers to studies that extract a large amount of quantitative information from medical imaging studies as a basis for characterizing a specific aspect of patient health. Radiomics models can be built to address a wide range of outcome predictions, clinical decisions, basic cancer biology, etc. For example, radiomics models can be built to predict the aggressiveness of an imaged cancer, cancer gene expression characteristics (radiogenomics), radiation therapy treatment response, etc. Technically, radiomics brings together quantitative imaging, computer vision/image processing, and machine learning. In this symposium, speakers will discuss approaches to radiomics investigations, including: longitudinal radiomics, radiomics combined with othermore » biomarkers (‘pan-omics’), radiomics for various imaging modalities (CT, MRI, and PET), and the use of registered multi-modality imaging datasets as a basis for radiomics. There are many challenges to the eventual use of radiomics-derived methods in clinical practice, including: standardization and robustness of selected metrics, accruing the data required, building and validating the resulting models, registering longitudinal data that often involve significant patient changes, reliable automated cancer segmentation tools, etc. Despite the hurdles, results achieved so far indicate the tremendous potential of this general approach to quantifying and using data from medical images. Specific applications of radiomics to be presented in this symposium will include: the longitudinal analysis of patients with low-grade gliomas; automatic detection and assessment of patients with metastatic bone lesions; image-based monitoring of patients with growing lymph nodes; predicting radiotherapy outcomes using multi-modality radiomics; and studies relating radiomics with genomics in lung cancer and glioblastoma. Learning Objectives: Understanding the basic image features that are often used in radiomic models. Understanding requirements for reliable radiomic models, including robustness of metrics, adequate predictive accuracy, and generalizability. Understanding the methodology behind radiomic-genomic (’radiogenomics’) correlations. Research supported by NIH (US), CIHR (Canada), and NSERC (Canada)« less
  • The rich data provided by high-resolution satellite imagery allow us to directly model geospatial neighborhoods by understanding their spatial and structural patterns. In this paper we explore an unsupervised feature learning approach to model geospatial neighborhoods for classification purposes. While pixel and object based classification approaches are widely used for satellite image analysis, often these approaches exploit the high-fidelity image data in a limited way. In this paper we extract low-level features to characterize the local neighborhood patterns. We exploit the unlabeled feature measurements in a novel way to learn a set of basis functions to derive new features. Themore » derived sparse feature representation obtained by encoding the measured features in terms of the learned basis function set yields superior classification performance. We applied our technique on two challenging image datasets: ORNL dataset representing one-meter spatial resolution satellite imagery representing five land-use categories and, UCMERCED dataset consisting of 21 different categories representing sub-meter resolution overhead imagery. Our results are highly promising and, in the case of UCMERCED dataset we outperform the best results obtained for this dataset. We show that our feature extraction and learning methods are highly effective in developing a detection system that can be used to automatically scan large-scale high-resolution satellite imagery for detecting large-facility.« less