Apoorva Safai¹, Sumeet Shinde¹, Manali Jadhav¹, Tanay Chougule¹, Abhilasha Indoria², Manoj Kumar², Vani Santosh², Shumyla Jabeen², Manish Beniwal², Subhash Konar², Jitender Saini², and Madhura Ingalhalikar^1
1Symbiosis International University, Pune, India, ²National Institute of Mental Health and Neurosciences, Bangalore, India

Adult Supratentorial Extraventricular Ependymoma (STEE) are rare neoplasms that are often misdiagnosed as high-grade gliomas (HGG) due to their similar radiological manifestation on MRI. However, the pathogenesis and treatment plan of ependymoma differs significantly from gliomas, and hence an early and accurate diagnosis is crucial. We propose a novel machine learning based diagnostic model that can accurately distinguish adult STEE from HGG subtypes using quantitative radiomic signatures from a multi-model MRI data. First order and texture based radiomic features, particularly from FLAIR, T2 and ADC, can capture intricate pathological variations and aid in accurate and differential diagnosis of STEE tumors.

Manali Balasaheb Jadhav¹, Richa Singh Chauhan², Priyanka Tupe Waghmare³, Archit Rajan⁴, Abhilasha Indoria², Jitender Saini⁵, Vani Santosh⁶, Madhura Ingalhalikar⁴, and Subhas Konar^7
1Symbiosis Center for Medical Image Analysis, Pune, India, ²Radiology, National Institute of Mental Health and Neuroscieces, Bengaluru, India, ³Symbiosis Institute on Technology, Pune, India, ⁴Symbiosis Centre for Medical Image Analysis, Pune, India, ⁵Radiology, National Institute of Mental Health and Neuroscieces, Pune, India, ⁶Neuropathology, National Institute of Mental Health and Neuroscieces, Bengaluru, India, ⁷Neurosurgery, National Institute of Mental Health and Neuroscieces, Bengaluru, India

H3K27M mutation in diffuse midline glioma is an independent predictor of overall survival however has a very poor prognosis. Identification of the mutation using conventional radiological analysis is complicated while the deep location of the tumors in the brain makes biopsy challenging with substantial risk of morbidity. To alleviate these issues, our work employs radiomics based machine learning framework to predict the H3K27M mutation from multi-modal MRI on 46 subjects. Results revealed 91% cross validation accuracy illustrating its future potential in clinical use.

Emma Bluemke¹, Ambre Bertrand¹, Kwun-Ye Chu^2,3, Nigar Syed², Andrew Murchison^3,4, Tessa Greenhalgh^3,5, Brian Burns⁶, Martin Craig⁷, Nia Taylor³, Ketan Shah^2,3, Fergus Gleeson^2,3, and Daniel Bulte^1
1University of Oxford, Oxford, United Kingdom, ²MRC Oxford Institute for Radiation Oncology, Department of Oncology, University of Oxford, Oxford, United Kingdom, ³Radiotherapy Department, Oxford University Hospitals NHS, Oxford, United Kingdom, ⁴Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom, ⁵University Hospital Southampton NHS FT, Southhampton, United Kingdom, ⁶GE Healthcare, Menlo Park, CA, United States, ⁷University of Nottingham, Nottingham, United Kingdom

As clinical studies adopt Oxygen-Enhanced MRI to assess its feasibility in human tumours, it is important for researchers to report any practical complications experienced. We report our practical experience with using both MOLLI and VFA T1 mapping for a recently completed OE-MRI clinical oncology study with the aim that sharing this is helpful to researchers planning to use OE-MRI in clinical oncology. Specifically, we report 4 elements: (1) difference in estimated T1 from each method used (MOLLI and VFA), (2) standard deviation within tumour ROIs, (3) OE-MRI response resulting from either method, and (4) artefacts and practical difficulties encountered.

SHUANG MENG¹, Ailian Liu¹, Lihua Chen¹, Qinhe Zhang¹, Qingwei Song¹, and Yunsong Liu^1
1The First Affiliated Hospital of Dalian Medical University, Dalian, China

It remains a challenge to diagnose the oligometastatic prostate cancer (PCa) due to the ambiguous definition of oligometastatic PCa. Previous studies had shown that dynamic contrast enhanced (DCE) - MRI imaging could be used to assess tumor aggressiveness. This study indicated that transfer constant (Ktrans) had the potential to assess the aggressiveness of PCa. And Ktrans combined with clinical characteristics (such as age and prostate specific antigen) had the higher diagnostic efficiency for oligometastatic and widely metastasis PCa.

Zhijun Geng¹, Yunfei Zhang², Chuanmiao Xie¹, and Yongming Dai^2
1Sun Yat-sen University Cancer Center, Guangzhou, China, ²Central Research Institute, United Imaging Healthcare, Shanghai, China

Susceptibility Weighted Imaging (SWI) has shown tremendous clinical significance for identifying the hepatic micro-structural abnormalities such as micro-bleeding, vascularity, nodule and so forth. Recent studies concluded that cytokeratin 19 (CK-19) is an important marker for prognostic prediction of hepatocellular carcinoma (HCC). We hypothesized that the neural network model can be established by means of extracting high throughput radiomics features from SWI images for noninvasively evaluating the CK-19 status with high accuracy. The results demonstrated that such deep learning based neural network model yielded excellent diagnostic performance for predicting the CK-19 status.

Xiaosheng Xu¹, Qinglei Shi², Weishuai Wang³, Jia Wei⁴, Li Yang⁴, Qian Xu⁴, and Gaofeng Shi^4
1The Fourth Hospital of Hebei Medical University, Shijiazhuang, China, ²MR Scientific Marketing，Siemens Healthcare, Beijing, China, ³MR Scientific Marketing，Siemens Healthcare, JINAN, China, ⁴The Fourth Hospital of Hebei Medical University, shijiazhuang, China

 In this study, we evaluated the scan time and image quality of a serious of gastric T2WI imaging, including conventional TSE with cartesian acquisition (TSE), BLADE-TSE, conventional HASTE sequence and a new TSE corrected HASTE sequence (TSE-HASTE). We found that the HASTE sequences can provide comparable image quality to TSE sequences, while is insensitive to motion and susceptibility artifacts. For patients with poor respiratory movement, HASTE sequence may be used as a good replacement to the TSE sequences, and TSE-HASTE sequence may provide a better image contrast compared to conventional HASTE.

Junyu Guo¹, Lauren Hinojosa¹, Yin Xi¹, Keith Husley¹, and Ivan Pedrosa^1
1Radiology, UT southwestern medical center, Dallas, TX, United States

Clear cell renal carcinoma cancer (ccRCC) is the most aggressive subtype among small renal masses. ccRCC identification can help in decision making between active surveillance and definitive intervention. Recently, a clear cell likelihood score (ccLS) using subjective interpretation of multiparametric MRI by radiologists was proposed. In this study, we investigate whether radiomics and deep learning (DL) technique can facilitate the prediction of ccRCC using T2-weighted images only. We compared the results of two different approaches, radiomics and DL, with the reported ccLS performance. Our results demonstrate that both radiomics and deep learning may provide useful information for identification of ccRCC.

Yuhui Liu¹, Ailian Liu¹, and Mingxiao Wang^1
1The First Affiliated Hospital of Dalian Medical University, Dalian, China, Dalian, China

This study aims to analyze the intrinsic relationship between imaging indicators and tumor biological behavior, and improve the clinical value of DCE-MRI.The correlation between DCE-MRI quantitative parameter values (Ktrans, Ve, Kep) and molecular biomarkers (Ki67,P53) was compared and explored in patients with rectal cancer.

 

Chen Chen¹, Fabao Gao¹, and Xiaoyue Zhou^2
1Department of Radiology, West China Hospital, Chengdu, China, ²MR Collaboration, Siemens Healthineers Ltd., Shanghai, China

Because of variations in severity and treatment methods of pilocytic astrocytoma, medulloblastoma, and ependymoma, accurate and specific diagnoses of the tumors are critical. Non-invasive diagnosis of posterior fossa tumors based on machine learning-based magnetic resonance imaging are being reported. However, conventional MRI, diffusion MRI, MR perfusion, and magnetic resonance spectroscopy have variable diagnostic values. We present here a meta-analysis of all the relevant published studies and conducted a large sample-size assessment concerning the diagnostic performance and potential covariates that could influence the diagnostic performance of machine learning.

Yikang Li^1,2,3, Zhan li Hu^1,2,3, Sen Jia^1,2,3, Wenjing Xu⁴, Zongyang Li^1,2,3, Hairong Zheng^1,2,3, Xin Liu^1,2,3, and Na Zhang^1,2,3
1Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, ²Key Laboratory for Magnetic Resonance and Multimodality Imaging of Guangdong Province, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, ³CAS key laboratory of health informatics, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, ⁴Faculty of Information Technology, Beijing University of Technology, Beijng, China

Nowadays, Magnetic Resonance Imaging (MRI) plays a pivotal role in gliomas diagnosis, analysis, and surgery planning. Nevertheless, the accuracy of MRI segmentation is enormously restricted by the quality of images. Therefore, we demonstrate a new method that can directly make segmentations from K space data. And the results show that our method achieves the state of the art.  

Andres Mauricio Arias-Lorza¹ and Natarajan Raghunand^1
1Moffitt Cancer Center, Tampa, FL, United States

In early-phase clinical trials, we observed a general decrease in tumor ADC following administration of PEGylated Recombinant Human Hyaluronidase PH20 (PEGPH20) as monotherapy. This decrease was related with reduction of tumor hyaluronan measured in biopsy samples. Reduction in ADC is suggestive of a decrease in tumor water content following hyaluronan depletion by PEGPH20.

Ye Li¹, Xulun Lu¹, Shifeng Tian¹, Jiazheng Wang², Zhiwei Shen², Qingwei Song¹, and Ailian Liu^1
1The First Affiliated Hospital of Dalian Medical University, Dalian, China, ²Philips Healthcare, Beijin, China

The accurate differential diagnosis of endometrial carcinoma and endometrial polyps always met many challenges. In this study, we evaluated the capability of differentiation of endometrial carcinoma from endometrial polyps with APT signal intensity (SI) and permeability parameters. The diagnostic efficiency is the highest using APT combined with Kep. APT and permeability parameters are potentially a promising and valuable method in differentiation of endometrial carcinoma from endometrial polyps.

Yifan Yuan¹, Qi Yue¹, Xiang Zou¹, Jiajun Cai¹, Ying-Hua Chu², Yi-Cheng Hsu², Patrick Alexander Liebig³, Hui Zhang⁴, He Wang⁴, Liang Chen¹, and Ying Mao^1
1Department of Neurosurgery, Huashan Hospital, Fudan University, Shanghai, China, ²MR Collaboration, Siemens Healthcare Ltd., Shanghai, China, ³Siemens Healthcare GmbH, Erlangen, Germany, ⁴Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, Shanghai, China

Preoperative assessment of histological grades and possible margin of glioma are challenges to neurosurgeons, which may also make a huge difference in therapeutic strategies. In this study, we aim to establish a grade-discrimination criterion via APT and explore the possible metastatic margin of diffuse glioma at 7T. Eleven patients underwent structural imaging and APT CEST imaging. Results of the APT were merged with the T1 image, which showed a promising indication of preoperative histological grades. Meanwhile, with the early exploration of metastatic margin, the APT CEST at 7 Tesla may significantly facilitate the surgical strategy and surgical excision extension.

Abdullah Bas¹, Banu Sacli-Bilmez¹, Ayca Ersen Danyeli^2,3, Cengiz Yakicier^3,4, M.Necmettin Pamir^3,5, Koray Ozduman^3,5, Alp Dincer^3,6, and Esin Ozturk-Isik^1,3
1Institute of Biomedical Engineering, Bogazici University, İstanbul, Turkey, ²Department of Medical Pathology, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ³Center for Neuroradiological Applications and Reseach, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁴Department of Molecular Biology and Genetics, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁵Department of Neurosurgery, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁶Department of Radiology, Acıbadem Mehmet Ali Aydinlar University, Istanbul, Turkey

Cramer-Rao lower bound (CRLB) is commonly employed as an exclusion criterion for bad quality MR spectra. This study aims to investigate the CRLB differences of the metabolites between isocitrate dehydrogenase (IDH) and telomerase reverse transcriptase promoter (TERTp) mutational subgroups in gliomas, and to assess the effects of CRLB threshold on their classifications. GPC, PCh, 2HG, and Ins were more reliably detected in IDH-mutant gliomas. Ins had higher CRLB values in TERTp-mutant than TERTp-wildtype gliomas. Different CRLB thresholds followed by zero-imputing had a small impact on the classification accuracies, but affected the choice of best features and classification algorithms.

Honglei Hu¹, Xixi Zhao¹, Qiming Wei², Chuyao Chen¹, Yuewei Huang¹, Yingjie Mei³, and Yikai Xu^1
1Medical Imaging Center, Nanfang Hospital, Southern Medical University, Guangzhou, China, ²Intervention, Traditional Chinese Medicine Hospital of Guangdong Province, Guangzhou, China, ³Philips Healthcare, Guangzhou, China

Rectal cancer is the third deadliest cancer in the world. The choice of its clinical treatment plan is closely related to clinical staging, which mainly depends on preoperative MR staging. Clinical work experience has found that the MR staging obtained by the existing MR technology is often less precise and often higher than the postoperative pathological staging. In order to solve this problem, our research has found that Amide Proton Transfer(ATP) imaging has a strong correlation with postoperative pathological staging, which can help clinicians to make a more accurate diagnosis of rectal cancer staging before surgery.

Banu Sacli-Bilmez¹, Ayca Ersen Danyeli^2,3, Murat Şakir Ekşi⁴, Kübra Tan⁵, Ozge Can⁶, Cengiz Yakicier⁷, M.Necmettin Pamir^3,4, Alp Dincer^3,8, Koray Özduman^3,4, and Esin Ozturk-Isik^1
1Institute of Biomedical Engineering, Bogazici University, Istanbul, Turkey, ²Department of Medical Pathology, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ³Center for Neuroradiological Applications and Reseach, Acıbadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁴Department of Neurosurgery, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁵Health Institutes of Turkey, Istanbul, Turkey, ⁶Department of Medical Engineering, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁷Department of Molecular Biology and Genetics, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁸Department of Radiology, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey

Meningiomas are a heterogeneous group of tumors that arise at the pachymeninges and tend to invade surrounding bone and rarely the brain. In this work, we studied the single-voxel ¹H-MRS correlates of aggressive biology (measured by WHO grade), peritumoral edema, brain invasion, bone invasion, skull base location in meningiomas. Higher Glyc and Ins+Glyc levels were observed in aggressive (WHO grade II or III) meningiomas. Higher Glu, GPC, and Glx levels were observed in hyperostotic tumors. The classification accuracies were 81.28% for tumor grade, 73.18% for hyperostosis detection, and 76.88% for skull base localization of meningiomas.

Zahra Alirezaei^1,2, Mohammadreza Nazemzadeh³, Masoud Hasanpour³, Alireza Amouheidari⁴, and Sajad Iraji^3
1Isfahan University of Medical Sciences, Isfahan, Iran (Islamic Republic of), ²Bushehr University of Medical Sciences, Bushehr, Iran (Islamic Republic of), ³Tehran University of Medical Sciences, Tehran, Iran (Islamic Republic of), ⁴Milad Hospital, Isfahan, Iran (Islamic Republic of)

To compare the sensitivity of alteration in MRS metabolites and MoCA and ACE-R cognitive tests in early and early-delayed post-radiation phases in detection of radiation-induced injury of Low Grade Glioam patients. The MRS metabolites, the ACE-R and MoCA, and the dosimetric parameters in the corpus callosum  were analyzed during RT and up to 6-month post-RT for 10 LGG patients. NAA/Cr and Cho/Cr declined significantly at least 3 months before detecting alterations in ACE and MoCA cognitive tests. Therefore, the MRS-based biomarkers may be more sensitive than the state of the art cognitive test tools in prediction of post-radiation cognitive impairments.

Andreas Korzowski¹, Nina Weckesser², Vanessa L Franke¹, Heinz-Peter Schlemmer², Mark E Ladd¹, Peter Bachert¹, and Daniel Paech^2
1Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany, ²Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany

The low spatial resolution of ³¹P MRSI of the human brain leads to a weighted mixture of specific spectral profiles from different tissue types in localized ³¹P spectra of individual voxels. Application of high-resolution ³¹P MRSI to volunteers and glioma patients at ultra-high B₀ yields a possibility to approach brain tissue-specific profiles, which could aid interpretation of observations from ³¹P MRSI. The presented high-quality ³¹P brain spectra from individual tissue types obtained at B₀ = 7T illustrate clear differences not only between healthy and tumor tissues, but also between different compartments within diseased tissue, i.e. contrast-enhanced regions and edema.

Abdullah BAS¹, Banu Sacli-Bilmez¹, Ayca Ersen Danyeli^2,3, Cengiz Yakicier^3,4, M.Necmettin Pamir^3,5, Koray Ozduman^3,5, Alp Dincer^3,6, and Esin Ozturk-Isik^1,3
1Institute of Biomedical Engineering, Bogazici University, Istanbul, Turkey, ²Department of Medical Pathology, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ³Center for Neuroradiological Applications and Reseach, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁴Department of Molecular Biology and Genetics, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁵Department of Neurosurgery, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, ⁶Department of Radiology, Acıbadem Mehmet Ali Aydinlar University, Istanbul, Turkey

Isocitrate dehydrogenase (IDH) and telomerase reverse transcriptase promoter (TERTp) mutations affect the clinical behavior and survival rate of diffuse gliomas. The detection of these mutations preoperatively is very critical for treatment planning. In this study, three different one dimensional convolutional neural network (1D-CNN) models were designed to identify IDH mutant (IDH-mut), TERTp mutant (TERTp-mut), and TERTp-only (IDH-wild type and TERTp-mut) gliomas based on proton magnetic-resonance spectroscopy (1H-MRS). The 1D-CNN models could identify IDH-mut, TERTp-mut, and TERTp-only gliomas with 94.11%, 76.92%, and 82.05% accuracies, respectively. This study showed the potential of deep-learning in predicting especially IDH-mutations in gliomas using 1H-MRS data.

Cornelius Cadrien^1,2, Sukrit Sharma¹, Philipp Lazen¹, Julia Furtner³, Alexandra Lipka^1,4, Eva Hečková¹, Lukas Hingerl¹, Stanislav Motyka¹, Stephan Gruber¹, Bernhard Strasser¹, Barbara Kiesel², Mario Mischkulnig², Matthias Preusser⁵, Thomas Roetzer⁶, Adelheid Wöhrer⁶, Michael Weber⁷, Christian Dorfer², Karl Rössler², Siegfried Trattnig^1,4, Wolfgang Bogner¹, Georg Widhalm², and Gilbert Hangel^1,2
1High-field MR Center, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, ²Department of Neurosurgery, Medical University of Vienna, Vienna, Austria, ³Division of Neuroradiology and Musculoskeletal Radiology, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, ⁴Christian Doppler Laboratory for Clinical Molecular MR Imaging, Vienna, Austria, ⁵Division of Oncology, Department of Inner Medicine I, Medical University of Vienna, Vienna, Austria, ⁶Division of Neuropathology and Neurochemistry, Department of Neurology, Medical University of Vienna, Vienna, Austria, ⁷Division of Medical Imaging and Nuclear Medicine, Medical University of Vienna, Vienna, Austria

We measured 38 HGG and LGG patients with a 15 min MR spectroscopic imaging sequence with an isotropic voxel size of 3.4 mm, covering a 64×64×39 matrix. After post-processing the measured spectra were analyzed statistically. Clinically segmented tumor regions showed statistically significant differences in metabolic ratios compared to normal-appearing white matter. Further analysis of metabolic hotspots with respect to tumor grade and IDH mutation status yielded no significant differences beyond myo-Inositol decreasing with tumor grade.