Ronald L. Wolf

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.

Clinical Neuroimaging Using Arterial Spin-Labeled Perfusion Magnetic Resonance Imaging

SummaryThe two most common methods for measuring perfusion with MRI are based on dynamic susceptibility contrast (DSC) and arterial spin labeling (ASL). Although clinical experience to date is much more extensive with DSC perfusion MRI, ASL methods offer several advantages. The primary advantages are that completely noninvasive absolute cerebral blood flow (CBF) measurements are possible with relative insensitivity to permeability, and that multiple repeated measurements can be obtained to evaluate one or more interventions or to perform perfusion-based functional MRI. ASL perfusion and perfusion-based functional MRI methods have been applied in many clinical settings, including acute and chronic cerebrovascular disease, CNS neoplasms, epilepsy, aging and development, neurodegenerative disorders, and neuropsychiatric diseases. Recent technical advances have improved the sensitivity of ASL perfusion MRI, and increasing use is expected in the coming years. The present review focuses on ASL perfusion MRI and applications in clinical neuroimaging.

Grading of CNS neoplasms using continuous arterial spin labeled perfusion MR imaging at 3 Tesla.

To differentiate glioma grade based on blood flow measured using continuous arterial spin labeled (CASL) perfusion MRI, implemented at 3 Tesla for improved signal‐to‐noise ratio (SNR) and spin labeling effect.

MRI identification of early white matter injury in anoxic–ischemic encephalopathy

Background: Anoxic–ischemic encephalopathy (AIE) affects the gray matter more than the white matter. Recent animal experiments suggest that the white matter is more sensitive to ischemia than previously thought. The authors describe the MRI findings in seven patients with AIE who demonstrate early preferential involvement of the white matter. Materials and methods: A retrospective case series study was performed, including seven patients with AIE who underwent MRI of the brain within 7 days of insult. Demographic information, type of insult, clinical examination findings, EEG findings, and clinical outcome were obtained. MRI studies were reviewed with specific attention to the cortex, deep gray matter, and the white matter structures. Mean apparent diffusion coefficient (ADC) was calculated in regions of interest placed in the cerebellar hemispheres, putamen, thalamus, splenium of corpus callosum, centrum semiovale, and medial frontal cortex. Results: The causes of AIE were cardiac arrhythmias in two patients, myocardial infarction in one, drug overdose in two, carbon monoxide poisoning in one, and respiratory failure and sepsis in one. The median time to MRI was 2.5 days. Symmetric areas of restricted diffusion were found in the periventricular white matter tracts (7/7 patients), the corpus callosum (6/7 patients), internal capsule (5/7 patients), and the subcortical association fibers (3/7 patients). ADC maps confirmed the restricted diffusion. Gray matter involvement was seen in three patients, and was more prominent on conventional imaging sequences compared with diffusion-weighted imaging. A subtle decrease in mean ADC was seen in cortex. Conclusions: Prominent, symmetric restricted diffusion can occur early after AIE in white matter, whereas gray matter involvement may be less prominent. Further studies involving a larger sample and serial imaging are required to confirm these preliminary findings.

Differentiation between glioblastomas and solitary brain metastases using diffusion tensor imaging

The purpose of this study is to determine whether diffusion tensor imaging (DTI) metrics including tensor shape measures such as linear and planar anisotropy coefficients (CL and CP) can help differentiate glioblastomas from solitary brain metastases. Sixty-three patients with histopathologic diagnosis of glioblastomas (22 men, 16 women, mean age 58.4 years) and brain metastases (13 men, 12 women, mean age 56.3 years) were included in this study. Contrast-enhanced T1-weighted, fluid-attenuated inversion recovery (FLAIR) images, fractional anisotropy (FA), apparent diffusion coefficient (ADC), CL and CP maps were co-registered and each lesion was semi-automatically subdivided into four regions: central, enhancing, immediate peritumoral and distant peritumoral. DTI metrics as well as the normalized signal intensity from the contrast-enhanced T1-weighted images were measured from each region. Univariate and multivariate logistic regression analyses were employed to determine the best model for classification. The results demonstrated that FA, CL and CP from glioblastomas were significantly higher than those of brain metastases from all segmented regions (p<0.05), and the differences from the enhancing regions were most significant (p<0.001). FA and CL from the enhancing region had the highest prediction accuracy when used alone with an area under the curve of 0.90. The best logistic regression model included three parameters (ADC, FA and CP) from the enhancing part, resulting in 92% sensitivity, 100% specificity and area under the curve of 0.98. We conclude that DTI metrics, used individually or combined, have a potential as a non-invasive measure to differentiate glioblastomas from metastases.

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.

Differentiation between Glioblastomas, Solitary Brain Metastases, and Primary Cerebral Lymphomas Using Diffusion Tensor and Dynamic Susceptibility Contrast-Enhanced MR Imaging

More on the eternal question: what can we use to differentiate preoperatively glioblastomas, metastases, and lymphomas? Here, the authors investigated whether diffusion tensor imaging and gadolinium perfusion studies could be used for this purpose. They evaluated 26 GBMs, 25 brain metastases, and 16 primary cerebral lymphomas with these techniques. Basically, GBMs showed lower fractional anisotropy and higher perfusion patterns. The best predictive data obtained were the apparent diffusion coefficients from enhancing tumor regions and the perfusion (cerebral blood volume) from the peritumoral regions. Although this is probably something that we all use on a daily basis, it is nice to see it reported in such an organized and careful fashion. BACKGROUND AND PURPOSE: Glioblastomas, brain metastases, and PCLs may have similar enhancement patterns on MR imaging, making the differential diagnosis difficult or even impossible. The purpose of this study was to determine whether a combination of DTI and DSC can assist in the differentiation of glioblastomas, solitary brain metastases, and PCLs. MATERIALS AND METHODS: Twenty-six glioblastomas, 25 brain metastases, and 16 PCLs were retrospectively identified. DTI metrics, including FA, ADC, CL, CP, CS, and rCBV were measured from the enhancing, immediate peritumoral and distant peritumoral regions. A 2-level decision tree was designed, and a multivariate logistic regression analysis was used at each level to determine the best model for classification. RESULTS: From the enhancing region, significantly elevated FA, CL, and CP and decreased CS values were observed in glioblastomas compared with brain metastases and PCLs (P < .001), whereas ADC, rCBV, and rCBVmax values of glioblastomas were significantly higher than those of PCLs (P < .01). The best model to distinguish glioblastomas from nonglioblastomas consisted of ADC, CS (or FA) from the enhancing region, and rCBV from the immediate peritumoral region, resulting in AUC = 0.938. The best predictor to differentiate PCLs from brain metastases comprised ADC from the enhancing region and CP from the immediate peritumoral region with AUC = 0.909. CONCLUSIONS: The combination of DTI metrics and rCBV measurement can help in the differentiation of glioblastomas from brain metastases and PCLs.

Multislice double inversion pulse sequence for efficient black‐blood MRI

Over the last several years there has been a rapidly growing interest in high‐resolution MRI of the vascular wall to assess the extent of atherosclerotic lesions. Vessels of particular clinical relevance are the carotid and coronary arteries. Currently, the preferred imaging sequence for these studies is a “black‐blood” technique based on the double‐inversion scheme to null the blood signal. A critical drawback of the black‐blood technique, however, has been its single‐slice nature, as there is only one point in time during the recovery of the blood magnetization from inversion at which the signal is completely nulled. Consequently, the total scan time can become prohibitively long, particularly when an imaging protocol includes several series of these datasets. In this work, a multiple‐slice double‐inversion technique is described that can reduce the scan time by a factor of two or more. It is demonstrated in vivo with examples from carotid and coronary arteries that one can acquire multiple slices with sufficient nulling of blood, following a single set of inversion pulses. Magn Reson Med 47:616–620, 2002.

Caudate blood flow and volume are reduced in HIV+ neurocognitively impaired patients

Objective: To evaluate the effects of HIV-associated neurocognitive impairment on caudate blood flow and volume. Methods: The authors performed continuous arterial spin labeled MRI on 42 HIV+ patients (23 subsyndromic and 19 HIV neurosymptomatic) on highly active antiretroviral therapy and 17 seronegative controls. They compared caudate blood flow and volume among groups. Results: A stepwise decrease in both caudate blood flow and volume was observed with increasing HIV-associated neurocognitive impairment. Compared with seronegative controls, baseline caudate blood flow was reduced in HIV+ neurosymptomatic patients (p = 0.001) with a similar decreasing trend for subsyndromic HIV+ patients (p = 0.070). Differences in caudate volume were observed only for neurosymptomatic HIV+ patients compared with controls (p = 0.010). A Jonckheere–Terpstra test for trends was significant for both caudate blood flow and volume for each of the three subgroups. Pearson product moment correlation coefficients were not significant between caudate blood flow and volume for each group. Conclusions: Decreasing trends in caudate blood flow and volume were associated with significantly increasing HIV-associated neurocognitive impairment (HNCI), with the greatest decreases observed for more severely impaired patients. However, reductions in caudate blood flow and volume were poorly correlated. Changes in residual caudate blood flow may act as a surrogate biomarker for classifying the degree of HNCI.

Arterial spin-labeling and MR spectroscopy in the differentiation of gliomas.

BACKGROUND AND PURPOSE: Noninvasive grading of gliomas remains a challenge despite its important role in the prognosis and management of patients with intracranial neoplasms. In this study, we evaluated the ability of cerebral blood flow (CBF)-guided voxel-by-voxel analysis of multivoxel proton MR spectroscopic imaging (1H-MRSI) to differentiate low-grade from high-grade gliomas. MATERIALS AND METHODS: A total of 35 patients with primary gliomas (22 high grade and 13 low grade) underwent continuous arterial spin-labeling perfusion-weighted imaging (PWI) and 1H-MRSI. Different regions of the gliomas were categorized as “hypoperfused,” “isoperfused,” and “hyperperfused” on the basis of the average CBF obtained from contralateral healthy white matter. 1H-MRSI indices were computed from these regions and compared between low- and high-grade gliomas. Using a similar approach, we applied a subgroup analysis to differentiate low- from high-grade oligodendrogliomas because they show different physiologic and genetic characteristics. RESULTS: Choglioma (G)/white matter (WM), GlxG/WM, and Lip+LacG/CrWM were significantly higher in the “hyperperfused” regions of high-grade gliomas compared with low-grade gliomas. ChoG/WM and Lip+LacG/CrWM were also significantly higher in the “hyperperfused” regions of high-grade oligodendrogliomas. However, metabolite ratios from the “hypoperfused” or “isoperfused” regions did not exhibit any significant differences between high-grade and low-grade gliomas. CONCLUSION: The results suggest that 1H-MRSI indices from the “hyperperfused” regions of gliomas, on the basis of PWI, may be helpful in distinguishing high-grade from low-grade gliomas including oligodendrogliomas.