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.

Machine learning (ML) integrated with medical imaging has introduced new perspectives in precision diagnostics of high-grade gliomas, through radiomics and radiogenomics. This has raised hopes for characterizing noninvasive and in vivo biomarkers for prediction of patient survival, tumor recurrence, and genomics and therefore encouraging treatments tailored to individualized needs. Characterization of tumor infiltration based on pre-operative multi-parametric magnetic resonance imaging (MP-MRI) scans may allow prediction of the loci of future tumor recurrence and thereby aid in planning the course of treatment for the patients, such as optimizing the extent of resection and the dose and target area of radiation. Imaging signatures of tumor genomics can help in identifying the patients who benefit from certain targeted therapies. Specifying molecular properties of gliomas and prediction of their changes over time and with treatment would allow optimization of treatment. In this article, we provide neuro-oncology, neuropathology, and computational perspectives on the promise of radiomics and radiogenomics for allowing personalized treatments of patients with gliomas and discuss the challenges and limitations of these methods in multi-institutional clinical trials and suggestions to mitigate the issues and the future directions.

BACKGROUND AND PURPOSE: Characterization of iron deposition associated with demyelinating lesions of multiple sclerosis and neuromyelitis optica has not been well studied. Our aim was to investigate the potential of ultra-high-field MR imaging to distinguish MS from neuromyelitis optica and to characterize tissue injury associated with iron pathology within lesions. MATERIALS AND METHODS: Twenty-one patients with MS and 21 patients with neuromyelitis optica underwent 7T high-resolution 2D-gradient-echo-T2* and 3D-susceptibility-weighted imaging. An in-house-developed algorithm was used to reconstruct quantitative susceptibility mapping from SWI. Lesions were classified as “iron-laden” if they demonstrated hypointensity on gradient-echo-T2*-weighted images and/or SWI and hyperintensity on quantitative susceptibility mapping. Lesions were considered “non-iron-laden” if they were hyperintense on gradient-echo-T2* and isointense or hyperintense on quantitative susceptibility mapping. RESULTS: Of 21 patients with MS, 19 (90.5%) demonstrated at least 1 quantitative susceptibility mapping–hyperintense lesion, and 11/21 (52.4%) had iron-laden lesions. No quantitative susceptibility mapping–hyperintense or iron-laden lesions were observed in any patients with neuromyelitis optica. Iron-laden and non-iron-laden lesions could each be further characterized into 2 distinct patterns based on lesion signal and morphology on gradient-echo-T2*/SWI and quantitative susceptibility mapping. In MS, most lesions (n = 262, 75.9% of all lesions) were hyperintense on gradient-echo T2* and isointense on quantitative susceptibility mapping (pattern A), while a small minority (n = 26, 7.5% of all lesions) were hyperintense on both gradient-echo-T2* and quantitative susceptibility mapping (pattern B). Iron-laden lesions (n = 57, 16.5% of all lesions) were further classified as nodular (n = 22, 6.4%, pattern C) or ringlike (n = 35, 10.1%, pattern D). CONCLUSIONS: Ultra-high-field MR imaging may be useful in distinguishing MS from neuromyelitis optica. Different patterns related to iron and noniron pathology may provide in vivo insight into the pathophysiology of lesions in MS.

Mutations in the isocitrate dehydrogenase enzyme are present in a majority of lower-grade gliomas and secondary glioblastomas. This mis-sense mutation results in the neomorphic reduction of isocitrate dehydrogenase resulting in an accumulation of the “oncometabolite” 2-hydroxyglutarate (2HG). Detection of 2HG can thus serve as a surrogate biomarker for these mutations, with significant translational implications including improved prognostication. Two dimensional localized correlated spectroscopy (2D L-COSY) at 7T is a highly-sensitive non-invasive technique for assessing brain metabolism. This study aims to assess tumor metabolism using 2D L-COSY at 7T for the detection of 2HG in IDH-mutant gliomas. Nine treatment-naïve patients with suspected intracranial neoplasms were scanned at 7T MRI/MRS scanner using the 2D L-COSY technique. 2D-spectral processing and analyses were performed using a MATLAB-based reconstruction algorithm. Cross and diagonal peak volumes were quantified in the 2D L-COSY spectra and normalized with respect to the creatine peak at 3.0 ppm and quantified data were compared with previously-published data from six normal subjects. Detection of 2HG was validated using findings from immunohistochemical (IHC) staining in patients who subsequently underwent surgical resection. 2HG was detected in both of the IDH-mutated gliomas (grade III Anaplastic Astrocytoma and grade II Diffuse Astrocytoma) and was absent in IDH wild-type gliomas and in a patient with breast cancer metastases. 2D L-COSY was also able to resolve complex and overlapping resonances including phosphocholine (PC) from glycerophosphocholine (GPC), lactate (Lac) from lipids and glutamate (Glu) from glutamine (Gln). This study demonstrates the ability of 2D L-COSY to unambiguously detect 2HG in addition to other neuro metabolites. These findings may aid in establishing 2HG as a biomarker of malignant progression as well as for disease monitoring in IDH-mutated gliomas.

BACKGROUND AND PURPOSE: Shutter-speed model analysis of dynamic contrast-enhanced MR imaging allows estimation of mean intracellular water molecule lifetime (a measure of cellular energy metabolism) and volume transfer constant (a measure of hemodynamics). The purpose of this study was to investigate the prognostic utility of pretreatment mean intracellular water molecule lifetime and volume transfer constant in predicting overall survival in patients with squamous cell carcinomas of the head and neck and to stratify p16-positive patients based upon survival outcome. MATERIALS AND METHODS: A cohort of 60 patients underwent dynamic contrast-enhanced MR imaging before treatment. Median, mean intracellular water molecule lifetime and volume transfer constant values from metastatic nodes were computed from each patient. Kaplan-Meier analyses were performed to associate mean intracellular water molecule lifetime and volume transfer constant and their combination with overall survival for the first 2 years, 5 years, and beyond (median duration, >7 years). RESULTS: By the last date of observation, 18 patients had died, and median follow-up for surviving patients (n = 42) was 8.32 years. Patients with high mean intracellular water molecule lifetime (4 deaths) had significantly (P = .01) prolonged overall survival by 5 years compared with those with low mean intracellular water molecule lifetime (13 deaths). Similarly, patients with high mean intracellular water molecule lifetime (4 deaths) had significantly (P = .006) longer overall survival at long-term duration than those with low mean intracellular water molecule lifetime (14 deaths). However, volume transfer constant was a significant predictor for only the 5-year follow-up period. There was some evidence (P < .10) to suggest that mean intracellular water molecule lifetime and volume transfer constant were associated with overall survival for the first 2 years. Patients with high mean intracellular water molecule lifetime and high volume transfer constant were associated with significantly (P < .01) longer overall survival compared with other groups for all follow-up periods. In addition, p16-positive patients with high mean intracellular water molecule lifetime and high volume transfer constant demonstrated a trend toward the longest overall survival. CONCLUSIONS: A combined analysis of mean intracellular water molecule lifetime and volume transfer constant provided the best model to predict overall survival in patients with squamous cell carcinomas of the head and neck.

This study aimed to investigate the potential of quantitative radiomic data extracted from conventional MR images in discriminating IDH-mutant grade 4 astrocytomas from IDH-wild-type glioblastomas (GBMs). A cohort of 57 treatment-naïve patients with IDH-mutant grade 4 astrocytomas (n = 23) and IDH-wild-type GBMs (n = 34) underwent anatomical imaging on a 3T MR system with standard parameters. Post-contrast T1-weighted and T2-FLAIR images were co-registered. A semi-automatic segmentation approach was used to generate regions of interest (ROIs) from different tissue components of neoplasms. A total of 1050 radiomic features were extracted from each image. The data were split randomly into training and testing sets. A deep learning-based data augmentation method (CTGAN) was implemented to synthesize 200 datasets from the training sets. A total of 18 classifiers were used to distinguish two genotypes of grade 4 astrocytomas. From generated data using 80% training set, the best discriminatory power was obtained from core tumor regions overlaid on post-contrast T1 using the K-best feature selection algorithm and a Gaussian naïve Bayes classifier (AUC = 0.93, accuracy = 0.92, sensitivity = 1, specificity = 0.86, PR_AUC = 0.92). Similarly, high diagnostic performances were obtained from original and generated data using 50% and 30% training sets. Our findings suggest that conventional MR imaging-based radiomic features combined with machine/deep learning methods may be valuable in discriminating IDH-mutant grade 4 astrocytomas from IDH-wild-type GBMs.