Xianqi Li¹, Bernhard Strasser¹, Kourosh Jafari-Khouzani², Daniel P Cahill³, Jorg Dietrich⁴, Tracy T Batchelor⁴, Martin Bendszus⁵, Ulf Neuberger⁶, Philipp Vollmuth⁶, and Ovidiu Andronesi^7
1Radiology, Massachusetts General Hospital, Charlestown, MA, United States, ²IBM Watson Health, Boston, MA, United States, ³Neurosurgery, Massachusetts General Hospital, Boston, MA, United States, ⁴Massachusetts General Hospital, Boston, MA, United States, ⁵Heidelberg University Hospital, Boston, MA, United States, ⁶Heidelberg University Hospital, Heidelberg, Germany, ⁷Massachusetts General Hospital,, Charlestown, MA, United States

We developed deep learning super-resolution MR spectroscopic imaging (MRSI) to map tumor metabolism in patients with mutant IDH glioma. A generative adversarial network (GAN) architecture comprised of a UNet neural network as the generator network and a discriminator network for adversarial training was employed to upsample MR spectroscopic imaging data with a factor of four. The preliminary results on simulated and in vivo data indicate that the proposed deep learning method is effective in enhancing the spatial resolution of metabolite maps which may better guide treatment in mutant IDH glioma patients. 

Robert R Edelman^1,2, Nondas Leloudas³, Jianing Pang⁴, Julian Bailes⁵, Ryan Merrell⁶, and Ioannis Koktzoglou^3,7
1Radiology, NorthShore University HealthSystem, EVANSTON, IL, United States, ²Feinberg School of Medicine, Northwestern University, Chicago, IL, United States, ³Radiology, NorthShore University HealthSystem, Evanston, IL, United States, ⁴Siemens Medical Solutions USA, Chicago, IL, United States, ⁵Neurosurgery, NorthShore University HealthSystem, Evanston, IL, United States, ⁶Medicine, NorthShore University HealthSystem, Evanston, IL, United States, ⁷Pritzker School of Medicine, University of Chicago, Chicago, IL, United States

We describe a novel class of steady-state pulse sequence called T1 Relaxation-Enhanced Steady-State (T₁RESS).  Several versions have been implemented including: (a) “bright blood” balanced T₁RESS (bT₁RESS), and (b) “dark blood” unbalanced T₁RESS (uT₁RESS).  There is also a two-echo Dixon version for fat/water separation.  In a brain tumor study, contrast-enhanced T₁RESS demonstrated a remarkable two-fold improvement in tumor-to-brain contrast as well as enhanced SNR compared with standard neuroimaging techniques, including MP-RAGE and T1 SPACE.  Our initial results suggest that T₁RESS has great potential for oncological imaging in the brain, and with further development in other organ systems as well.

jun zhang^1,2, Hong Peng¹, Yu-Lin Wang¹, De-Kang Zhang¹, and Lin Ma^1
1Radiology, Chinese PLA general hospital, BeiJing, China, ²radiology, the sixth center of Chinese PLA general hospital, BeiJing, China

The IDH status has been reported as major prognostic factors for glioma behavior. Thus, noninvasively detecting molecular subtypes before surgery is important for predicting the outcome and choosing therapy. In this study, using machine-learning algorithms, the accurate prediction of IDH subtype was achieved for diffuse gliomas via noninvasive MR imaging, including ADC values and tumor morphologic features, and it is worth mentioning ADC measurements applied are available in clinical workstations. Furthermore, to our knowledge, no previous attempts have been made to use different machine-learning methods to build a suitable model to predict the IDH status for WHO grade II-IV gliomas.

Qihao Zhang¹, Gloria Chia-Yi Chiang², Thanh Nguyen², Pascal Spincemaille², and Yi Wang^1
1Cornell University, New York, NY, United States, ²Weill Cornell Medical College, New York, NY, United States

We purpose to test the performance of glioblastoma grading using several perfusion parameters: QTM velocity U, volume transfer constant K^trans, and blood volume fraction V_e. For K^trans and V_e calculated using extended Toft’s model, two arterial input function are used: one sampled from Carotid artery and one sampled from the feeding artery near the tumor. The result shows that U has the best classification accuracy comparing with other parameters.

Pegah Askari¹, Ivan E Dimitrov¹, Michael Levy¹, Toral R Patel¹, Edward Pan¹, Bruce E Mickey¹, Craig R Malloy¹, Elizabeth A Maher¹, and Changho Choi^1
1University of Texas Southwestern Medical Center, Dallas, TX, United States

In proton MRS, the 2.25 ppm resonance of 2-hydroxyglutarate (2HG), which gives the largest signal in many acquisition methods, is overlapped with the lipid resonance at 2.25 ppm, complicating evaluation of 2HG in necrotic tumors with elevated lipids. We propose a novel approach for separation of the 2HG and lipid overlapped signals in spectral fitting. We developed new lipid basis sets and tested on proton MRS (PRESS TE 97ms at 3T) data from 43 glioma patients. LCModel fitting using the new lipid basis sets resulted in complete distinction between IDH mutation and wildtype (accuracy, sensitivity, and specificity all unity).

Xiaoping Zhu¹, Daniel Lewis², Ka-Loh Li¹, Sha Zhao³, Timothy Cootes¹, Andrew King², David Coope², and Alan Jackson^3
1DIIDS, University of Manchester, Manchester, United Kingdom, ²Neurosurgery, Salford Royal NHS Foundation Trust, Manchester, United Kingdom, ³University of Manchester, Manchester, United Kingdom

Accurate derivation of a vascular input function (VIF) is essential for pharmacokinetic modeling of brain DCE-MRI data. In this patient study we compared VIFs extracted from either the internal carotid artery (ICA) and its branches or the superior sagittal sinus (SSS).  We demonstrate that plasma gadolinium based contrast-agent (GBCA) concentration-time curves during the 1^st-pass of the bolus are equivalent in shape in both the ICA and SSS; and that compared to large intracranial arteries the SSS is able to provide a superior global VIF, with lower noise, higher peak-amplitude and demonstrably greater sensitivity to inter-individual changes in plasma GBCA concentration.

Alan Finkelstein¹, Arun Venkataraman², Madalina Tivarus³, Md Nasir Uddin⁴, Jianhui Zhong^1,2,3, Giovanni Schifitto^3,4, Michael Milano⁵, Michelle Janelsins⁶, and Sara Hardy^4,5
1Biomedical Engineering, University of Rochester, Rochester, NY, United States, ²Physics and Astronomy, University of Rochester, Rochester, NY, United States, ³Imaging Sciences, University of Rochester, Rochester, NY, United States, ⁴Neurology, University of Rochester, Rochester, NY, United States, ⁵Radiation Oncology, University of Rochester, Rochester, NY, United States, ⁶Surgery, University of Rochester, Rochester, NY, United States

Patients receiving partial brain radiotherapy (RT) experience cognitive impairment. Determining what structures and networks in the brain are susceptible to RT will help to elucidate mechanisms of cognitive impairment in the setting of RT. In this abstract, we examine structural and functional changes before and 6 months post RT in patients with lower grade brain tumors. Further, graph theory analysis of functional connectivity (FC) data was used to investigate network topology. We found that  subcortical volume and local graph theory metrics were significantly correlated with RT dose. We also propose a putative link between volumetric changes and network topology.

Qing Fu^1,2, Qi-guang Cheng^1,2, Xiao-yong Zhang³, Xiang-chuang Kong^1,2, Ding-xi Liu^1,2, Yi-hao Guo⁴, John Grinstead⁵, Zi-qiao Lei^1,2, and Chuan-sheng Zheng^1,2
1Department of Radiology, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China, ²Hubei Province Key Laboratory of Molecular Imaging, Wuhan, China, ³Centers for Biomedical Engineering, University of Science and Technology of China, Hefei, China, ⁴MR Collaboration, Siemens Healthcare Ltd., Guangzhou, China, ⁵Siemens Medical Solutions USA, Inc., Portland, OR, United States

Three dimensional (3D) T1-weighted sequences are commonly used for imaging of contrast-enhanced brain tumors, but enhancement of small vessels may mimic imaging of focal lesions, thereby hindering accurate diagnosis. This study validated that when compared with PETRA and conventional MPRAGE, DANTE-SPACE detects more small metastases with better CNR between lesions and surrounding parenchyma, and may be useful for differential tumor diagnosis because of its blood signal suppression abilities; PETRA could achieve similar detection for brain tumors and quieter acoustic noise level when compared with MPRAGE.

Augusto Lio M. Goncalves Filho^1,2, John Conklin^1,2, Chanon Ngamsombat³, Stephen F. Cauley², Wei Liu⁴, Daniel N. Splitthoff⁵, Wei-Ching Lo⁶, John E. Kirsch¹, Pamela W. Schaefer¹, Otto Rapalino¹, and Susie Y. Huang^1,2
1Department of Radiology, Massachusetts General Hospital, Boston, MA, United States, ²Department of Radiology, Athinoula A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States, ³Department of Radiology, Siriraj Hospital, Bangkok, Thailand, ⁴Siemens Shenzhen Magnetic Resonance Ltd., Shenzhen, China, ⁵Siemens Healthcare GmbH, Erlangen, Germany, ⁶Siemens Medical Solutions Inc., Boston, MA, United States

We compared a highly accelerated post-contrast Wave-CAIPI 3D T1 MPRAGE sequence to the standard 3D T1 MPRAGE in 80 patients undergoing imaging evaluation of intracranial enhancing lesions on 3T MRI. Two neuroradiologists performed a head-to-head analysis to assess the images and found that post-contrast 3D Wave-T1 MPRAGE was up to 2x faster than the standard sequence with no significant difference in the detection of enhancing lesions. The application of highly accelerated Wave-T1 MPRAGE may improve use of MRI resources while reducing patient anxiety and being less prone to motion artifacts.

Shin Tai Chong¹, Jianping Song², Kuan-Tsen Kuo³, Yu-Ting Ko³, Sanford PC Hsu⁴, Jinsong Wu², and Ching-Po Lin^1
1Institute of Neuroscience, National Yang-Ming University, Taipei, Taiwan, ²Department of Neurosurgery, Huashan Hospital, Shanghai Medical College, Fudan University, Shanghai, China, ³ABC Solution Co., Ltd, Shanghai, China, ⁴Department of Neurosurgery, Neurological Institute, Taipei Veterans General Hospital, Taipei, Taiwan

DiffusionGo is an in-house developed software specially designed for neurosurgeons, which integrated a reliable pre-processing pipeline, fully automatically fiber reconstruction algorithm, and multimodalities imaging information to achieve easy modeling and precision surgery. 

Yan Tan¹, Wenwei Shi², Xiaochun Wang¹, and Hui Zhang^1
1Department of Radiology, First Hospital of Shanxi Medical University, Taiyuan, China, ²Department of Radiology, Zhongda Hospital, Southeast University, Nanjing, China

To evaluate the diagnostic performance of DKI in differentiating glioma recurrence from pseudoprogression and the combined value of DKI and DSC MRI parameters. Tumor recurrence was confirmed to be associated with more tumor angiogenesis, greater nuclear atypia, and increased cell density, whereas pseudoprogression is characterized by radiation-induced vascular changes leading to vasodilation, edema, and increased capillary permeability. These resulted in a more complex structure of recurrent tumor than pseudoprogression. It is concluded DKI and DSC MRI may serve as imaging biomarker of treatment response by characterizing the heterogeneity of the microenvironment and newly formed immature blood vessels.

Yi Sui¹, Myung-Ho In¹, Salomon Cohen-Cohen², Erin Gray¹, Kevin Glaser¹, Matt A. Bernstein¹, Jamie J Van Gompel², Richard L. Ehman¹, John Huston III¹, and Ziying Yin^1
1Radiology, Mayo Clinic, Rochester, MN, United States, ²Neurologic Surgery, Mayo Clinic, Rochester, MN, United States

Preoperatively knowing if a pituitary adenoma is adhering to the optic chiasm would help to predict the risk of surgically induced vision loss. MR-elastography based slip interface imaging has been successfully developed to predict tumor adhesion using normalized octahedral shear strain (NOSS). This pilot study demonstrates the feasibility of acquiring NOSS at ~1mm high resolution using a newly developed distortion-free EPI-MRE technique on a high-performance compact 3T system. With further development, this technique will have great potential to identify adhesion between pituitary adenomas with the adjacent optic chiasm.  

Xiaoxin Li¹, Qingwei Song¹, Ailian Liu¹, Lizhi Xie², and Yanwei Miao^1
1First Affiliated Hospital of Dalian Medical University, Dalian, China, ²GE Healthcare, MR Research China, Beijing, China

To evaluate the application value of texture analysis based on conventional MRI for predicting the cell proliferation status of glioma. Patients with glioma were divided into two groups by Ki-67, group 1 was Ki-67 less than or equal to 10% and group 2 was Ki-67 more than 10%. The texture feature value of T1WI, T2WI, and contrasted T1WI were obtained by Omni-Kinetics software. The texture features of three sequences were significantly different between two groups, and uniformity and correlation showed better diagnostic efficiency. So the texture analysis based on conventional MRI is useful to predicting cell proliferation status of glioma. 

Tao Gong¹, Liangjie Lin², and Yihang Yang^3
1Shandong Medical Imaging Research Institute, Jinan, China, ²Philips Healthcare, Beijing, China, ³Shandong provincal hospital, Jinan, China

The aim of this study was to detect the isocitrate dehydrogenase 1 (IDH 1) genotype in low grade gliomas using diffusion kurtosis imaging (DKI). The results indicated that  IDH 1 wild-type gliomas have higher MK and lower MD values in tumor foci compared with mutant gliomas, and MK increased and MD decreased significantly in normal-appearing perilesional white matter (pWM) in all gliomas. Interestingly, lower FA in pWM than contralateral normal-appearing white matter was only seen in wild-type gliomas, while no difference was found in mutant-gliomas. These findings indicated that DKI were sensitive to differentiate IDH 1 genotypes in low-grade gliomas.

Nadya Pyatigorskaya^1,2,3, Arnaud Pellerin¹, Maya Kalife⁴, Marc Bertaux⁵, Marine Soret⁵, Didier Dormont¹, and Aurélie Kas^5
1Neuroradiology, Pitié-Salpêtrière Hospital, APHP, Sorbonne universite, Paris, France, ²UMR S 1127, CNRS UMR 722, ICM, Sorbonne Universités, Paris, France, ³CENIR, ICM, Sorbonne Universités, Paris, France, ⁴CENIR, ICM, Paris, France, ⁵Nuclear Medecine, Pitié-Salpêtrière Hospital, APHP, Sorbonne universite, Paris, France

We aimed at investigating the methods based on coupling cerebral perfusion (ASL) and amino-acid metabolism (18F-DOPA-PET) measurements to evaluate the diagnostic performance of simultaneous PET/MRI in glioma follow-up. Tumour isocontour maps of 18F-DOPA-PET and ASL T-maps were created using SPM and metabolic/perfusion abnormalities were evaluated with the asymmetry index z-score. SPM map analysis of significant-size clusters and semi-quantitative PET and ASL map evaluation were performed. The tumour isocontour maps and T-maps showed the highest specificity (100%) and sensitivity (94.1%) for ASL and 18F-DOPA analysis, allowing to achieve high diagnostic performance in differentiating between progression and pseudo-progression in treated gliomas.

Zahra Shams¹, Sarah M. Jacobs¹, Jan W. Dankbaar¹, Changho Choi², Dennis W.J. Klomp¹, Jannie P. Wijnen¹, and Evita C. Wiegers^1
1Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands, ²Advanced Imaging Research Center, University of Texas Southwestern Medical Center, Dallas, TX, United States

2HG is a valuable biomarker to detect mutational status of glioma patients. Therefore, it is of great value to be widely integrated into clinical practice. This study investigates the performance of along echo time (PRESS 97ms) ¹H MRS protocol at 3Tesla to be used in routine clinical glioma imaging. 

Chan Hong Moon¹, Krystof Bankiewicz², Paul Larson², Adrian Kells³, Alastair J. Martin², Stephen Mancuso⁴, and Mark Richardson^5
1University of Pittsburgh, Pittsburgh, PA, United States, ²University of California San Francisco, San Francisco, CA, United States, ³Voyager Therapeutics Inc., Cambridge, MA, United States, ⁴UPMC, Pittsburgh, PA, United States, ⁵Massachusetts General Hospital, Boston, MA, United States

Neurosurgical planning/targeting/monitoring by MRI requires high contrast and excellent anatomical feature. Inversion recovery MRI, e.g., MPRAGE is known to be superior to T1w SPGR in making high contrast of WM vs. GM. However, it is not optimized for putamen/globus pallidus, i.e., important target nuclei in PD surgery, particularly iMRI scanner 1.5T (not 3T). In this study, we developed new dual-contrast MPRAGE and maximized T1 contrast in putamen as well as good anatomy of whole brain without additional acquisition compared to conventional MPRAGE. The proposed methods were applied to 1.5T iMRI during neurosurgical procedure of PD patients.

Uzay E Emir^1,2, Natalie E Voets³, Sarah E Larkin⁴, Nick De Pennington⁴, Puneet E Plaha⁴, Richard E Stacey⁴, James E Mccullagh⁴, Christopher J Schofield⁴, Stuart E Clare³, Peter Jezzard³, Thomas Cadoux-hudson⁴, and Olaf E Ansorge^4
1School of Health Sciences, Purdue University, West Lafayette, IN, United States, ²Weldon School of Biomedical Engineering, Purdue University, West Lafayette, IN, United States, ³Wellcome Centre for Integrative Neuroimaging, Nuffield Dept of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom, ⁴University of Oxford, Oxford, United Kingdom

Given the rapid pace of discoveries in glioma genomics and metabolomics, it is likely that improvements in diagnostic imaging using UHF MRS can be made by systematically reanalyzing previously generated data in light of new knowledge. In this study, we tested this by reanalyzing existing UHF MRS spectra using a refined basis set that included the potential oncometabolite cystathionine

Chuanke Hou¹, Guanzhong Gong¹, Weiqiang Dou², and Yong Yin^1
1Department of Radiation Physics, Shandong Cancer Hospital and Institute, Shandong First Medical University and Shandong Academy of Medical Sciences, Jinan, China, ²GE Healthcare, MR Research China, Beijing, China

In this work, brain metastases (BM) were divided into multiple sub-volumes based on the cerebral blood flow (CBF) distribution derived by 3D arterial spin labeling (ASL) perfusion imaging. Due to the previously reported strong relationship, low CBF area was defined as the hypoxia area and selected as the region of interest (ROI). On this ROI, we further investigated the behaviors of conventional Intensity-modulated Radiotherapy (IMRT) plan, simultaneous integrated boost IMRT plan with and without constraints by increasing targeted doses, aiming to provide a new reference model for individualized radiotherapy for BM patients.

Sean P Devan¹, Xiaoyu Jiang¹, Guozhen Luo², Jingping Xie¹, Zhongliang Zu¹, Ashley M Stokes³, Austin N Kirschner², John C Gore¹, and Junzhong Xu^1
1Institute of Imaging Science, Vanderbilt University Medical Center, Nashville, TN, United States, ²Radiation Oncology, Vanderbilt University Medical Center, Nashville, TN, United States, ³Keller Center for Imaging Innovation, Barrow Neurological Institute, Phoenix, AZ, United States

It remains a challenge to differentiate between recurrent tumors from radiation induced necrosis in the brain. One promising approach is to combine complementary information on different scales obtained using multi-parametric MRI to comprehensively characterize tissue histopathological properties. In this preliminary study, we performed six multi-parametric MRI acquisitions including APT (probing mobile proteins), NOE (mobile macromolecules), qMT (macromolecules), ADC (cell density), SSIFT (cell size), and DSC (perfusion) to differentiate 9L gliosarcoma from radiation necrosis in animal models. The results suggest APT and SSIFT provide the best discrimination of different tissue types, but a larger sample size is required for further validation.