eRapport

Development of New Imaging Modalities Enable Tailored Therapies in IDH Mutated Gliomas

Prosjekt
Prosjektnummer
2018047
Ansvarlig person
Morteza Esmaeili
Institusjon
Akershus universitetssykehus HF
Prosjektkategori
Forskerstipend
Helsekategori
Cancer, Neurological
Forskningsaktivitet
4. Detection and Diagnosis
Rapporter
2023 - sluttrapport
Tailoring medical approaches for patients with brain tumors requires advanced imaging techniques. Molecular imaging of the brain through magnetic resonance spectroscopic imaging (MRSI) allows precise measurement of neurochemicals in the body. Proton MRSI excels at identifying metabolites associated with altered cellular pathways in cancer. The primary goals of this project include exploring advanced technologies to improve and expand MRSI for brain tumors. We focus on in vivo detection of 2-hydroxyglutarate (2-HG), the metabolite that accumulates in brain tumor tissues harboring isocitrate dehydrogenase (IDH) mutation. Our exploration extends to alternative advanced imaging methods in brain studies, incorporating MR spectroscopy and MRSI in dementia, skeleton and cardiac muscles studies. Throughout sub-projects, we demonstrated the significant potential of MRSI, especially in brain tumor examinations, emphasizing its seamless integration with other advanced MRI and PET methods. In different project phases, we highlighted the necessity for methodological development, covering acquisitions involving sequence program modification or shim array coil, as well as data analysis utilizing artificial intelligence. Our successful demonstrations encompassed the development of innovative imaging approaches like 2-HG MRSI at 7 Tesla, providing valuable tools for tailoring personalized patient treatment and conducting follow-up assessments. We also explored the application of data-driven artificial intelligence on conventional MRI data, investigating its potential to enhance diagnostic performance in brain tumor examinations. The knowledge gained in advanced MRSI, spectroscopic data analysis, image analysis, and AI algorithm development contributed to studies in partner hospitals in southeast Norway, resulting in publications across the fields of Alzheimer’s disease, functional MRI analysis, and cardiac MR spectroscopy. In Norway, we have one 7 Tesla MRI and several new 3 Tesla MRI systems that have been recently installed. Moreover, new hospitals are under construction in the southeast region. Most of these MRI systems are not equipped with advanced sequence programs, particularly MRSI protocols. While PET provides excellent molecular information, establishing Metabolite MRI protocols, in the context of advanced MRSI acquisition, can significantly contribute to the diagnostic and treatment monitoring of individuals with neuropathological diseases and cancer. This is because they can be run in the same scanning session (offering more convenience and economic sense) using the same equipment, and it is radiation-free. Also, MRSI data will be easily co-registered and fused with anatomical MRI data, which is favorable for interpretations. The knowledge gained in our study can significantly contribute to facilitating these opportunities by collaborating with the radiology departments, disseminating findings and knowledge, and educating and training researchers and doctoral students.

NEI

2022
Precisely classifying gliomas (common primary malignancies in the brain) is crucial for accurate diagnosis and clinical management. Metabolic brain imaging using magnetic resonance spectroscopic imaging (MRSI) enables neurochemical quantification in vivo. MRSI can detect metabolites from several cellular altered metabolic pathways.Lower-grade gliomas exhibit several tumor-specific molecular and genomic alterations despite the pathological characterization. An early mutation in the isocitrate dehydrogenase (IDH) genes has been frequently detected in a subgroup of patients with lower-grade gliomas. Several studies have demonstrated that IDH-mutated gliomas are associated with a better prognosis than IDH wild-type gliomas. Findings in distinct clinical behaviors of IDH-mutated gliomas have led to an updated WHO grading sub-classification of gliomas since 2016. This project aims to develop advanced MRSI technologies and processing tools to detect important neurochemicals in the brain. Also, we aim to develop artificial intelligence- (AI) based methods to provide spectroscopy data processing and classification. In 2022, we examined a modified version of densenet deep learning to develop and implement AI-based MR data analysis tools. We implemented this AI model on functional MRI data to predict behavior scores from brain functional connectivity fingerprints. To study cancer and tumor microenvironment, we may examine multimodal approaches if we could improve the efficiency of data acquisition. For example, a combination of MRSI with fMRI data and positron emission tomography (PET) imaging can significantly contribute to precise medicine of brain tumor examinations. We have published the results from some the above-mentioned investigations in international conferences, a book chapter, and research articles. Also, we emphasized the role of MR spectroscopy in other areas of clinical examination, such as cardiac MRS published in 2022. Task 1. Three-dimensional whole-brain MRSI and fully-automated and reproducible data analysis can significantly extend the clinical utility of the MRSI. Data analysis and interpretation require extensive pre-processing and expertise and often discourage the end-users from using the technique in clinical neurology. We aim to develop DL models to process MRS raw data. The network can extract low-dimensional features of the spectra and reconstruct the frequency domain. This work will continue in 2023. Task 2: PI contributed to writing a book chapter preparation entitled: “MR-Derived Biomarkers for Cancer Characterization. I: Biomarkers of the Tumor Microenvironment”. Springer. Task 3. To develop an artificial intelligence model predicting individual behavior scores from functional connectivity fingerprints. The scores include episodic memory and working memory. We believe the methodological development from the study can contribute to neuro-pathology investigations such as Alzheimer’s disease. We will continue our investigations in 2023. Task 4: Investigate the fusion of MRSI and PET data to examine the diagnostic power of these imaging modalities in characterizing glioma subtypes. This is an ongoing project in 2023.

NO

2021
A precise classification of gliomas (common primary malignancies in the brain) is crucial for accurate diagnosis and clinical management. Metabolic imaging of the brain using magnetic resonance spectroscopic imaging (MRSI) enables neurochemical quantification in vivo. MRSI can detect metabolites from several cellular altered metabolic pathways.Lower-grade gliomas exhibit several tumor-specific molecular and genomic alterations despite the pathological characterization. An early mutation in the isocitrate dehydrogenase (IDH) genes has been frequently detected in a subgroup of patients with lower-grade gliomas and associated with a better prognosis than the IDH wild-type gliomas. Findings in distinct clinical behaviors of IDH mutated gliomas have led to an updated WHO grading sub-classification of gliomas since 2016. This project aims to develop advanced MRSI technologies to detect important neurochemicals in the brain. Also, we aim to develop artificial intelligence- (AI) based methods to provide automatic spectroscopy data processing. As a class of machine learning, deep learning (DL) methods can identify patterns in large datasets. In 2021, we mainly examined different methodological approaches to develop and implement AI-based MR data analysis tools. Diffusion tensor imaging and functional MRI provide valuable neurological information of the brain. We conducted two functional imaging studies. These investigations were published in national and international conferences and research articles. Task 1. This study examined an explainable method to evaluate the high-level features of deep learning methods in tumor localization. We developed artificial intelligence-based algorithms to characterize lower-grade gliomas based on anatomical MR images. The proposed training evaluation may improve human-machine interactions and assist in selecting an optimal training scheme for clinical questions and machine learning progress. The results from this task were published in the Journal of Personalized Medicine. Task 2. Three-dimensional whole-brain MRSI and fully-automated and reproducible data analysis can significantly extend the clinical utility of the MRSI. Data analysis and interpretation require extensive pre-processing and expertise and often discourage the end-users from using the technique in clinical neurology. We aim to develop DL models to process MRS raw data. The network can extract low dimensional features of the spectra and reconstruct the frequency domain. The task's preliminary results were presented at the Nordic Artificial Intelligence conference. This work continues in 2022. Task 3. Dementia is one of the leading public health concerns. Although Alzheimer's disease (AD) is the most common dementia diagnosis among older patients, some patients have behavioral symptoms. Therefore, it is essential to achieve an exact diagnosis to provide the best possible treatment and perform research. Diffusion tensor imaging (DTI) measures the diffusion of water through the brain tissue and can visualize and measure the integrity of white matter tracts. Fraction anisotropy (FA) is a valuable connectivity metric derived from DTI. This task aimed to investigate whether FA measures can accurately find patients with behavioral symptoms within a group of AD patients. The results from this task were published in Acta Radiologica. Task 4. The study aims to develop an artificial intelligence platform to predict individual behavior from individual functional connectivity profiles, such as episodic memory scores. The secondary aim is to identify the features of the deep convolutional neural network model that contributed to predictions of EM. The preliminary results of this study were submitted to conferences: human brain mapping and ISMRM2022. We will continue our investigations in 2022.

NO

2020
Precision medicine of gliomas requires advanced imaging modalities. Molecular imaging of the brain using magnetic resonance spectroscopic imaging (MRSI) enables neurochemical quantification in vivo. Proton MRSI can detect metabolites from several cellular altered metabolic pathways in cancer.A precise classification of gliomas (common primary malignancies in the brain) is crucial to ensure an accurate diagnosis and clinical management. Despite the pathological characterization, lower-grade gliomas exhibit several tumor-specific molecular and genomic alterations. An early mutation in the isocitrate dehydrogenase (IDH) genes has been frequently detected in a subgroup of patients with lower-grade gliomas and associated with a better prognosis than the IDH wild type gliomas. MRSI offers a significant contribution in gliomas’ diagnostics and response evaluation to novel targeted therapies. A key clinical application of MRSI has emerged in neuro-oncology and neurosurgery by imaging the oncometabolite D-2-hydroxyglutarate (2-HG) in mutant IDH gliomas, that hence may be used for glioma subtyping. This project aims to develop advanced MRSI technologies to detect the 2-HG metabolite. The strong magnetic field (B0) requires more technical adjustments. It is difficult to obtain a homogeneous B0 field uniformly across the brain, especially at ultra-high-field (7 Tesla). Ultra-high-field provides increased sensitivity for MRSI examinations. However, a homogeneous B0 field is critical for high-quality MRSI data. We successfully designed, developed, and implemented two parallel technical tasks to tackle the challenges. The results from these investigations were published in 2020. Task 1. We developed a novel k-space trajectory to incorporate a 3D fast MRSI for human brain metabolite imaging at 7 Tesla. The spiral out-in trajectory increased the signal-to-noise ratio and data sampling efficiency significantly. With improved data collection, we could quantify more voxels in the brain. The results from this task were published in the Journal of Magnetic Resonance in Medicine. The results also attracted attention in the community. The editorial board invited us to submit a banner and featured research figure. Task 2. For the first time, we examined the performance of a shim array coil on a human 7 Tesla MRI system. The significant findings from this project were multiple improvements on critical parameters of the MRSI, including signal-to-noise ratio, the linewidth of the metabolites signals, increased number of usable voxels in the brain region of interest, and reduced signal fitting errors. The project was constructed on a multidisciplinary approach that expertise from different backgrounds and collaboration with international partners. The results from this project were published in Nature Scientific Reports. Task 3. In a collaborative project with researchers at the University of Cambridge and the NTNU, the PI performed MRS on samples acquired from breast cancer models to investigate metabolic response on a novel treatment developed at the NTNU. The metabolite quantification results demonstrated the sensitivity of metabolic quantification in differential evaluation of the treatment. Task 4. We developed artificial intelligence-based algorithms to characterize lower-grade gliomas based on anatomical MR images. This study uses an explainable method to evaluate the high-level features of deep learning methods in tumor localization. The proposed training evaluation may improve human-machine interactions and assist in selecting an optimal training scheme for clinical questions and machine learning progress. We have submitted the preliminary results to the 2021 ISMRM Annual Meeting.

No

2019
The purpose of the project is to investigate the role of molecular imaging in more accurate diagnosis, prognosis, and response to therapy assessment in cancer. MRSI provides non-invasive in vivo metabolic profiling of neurologic pathologies such as tumors. It detects metabolites from several pathways of importance in cancer.Advanced and optimized versions of MRSI enable detection of 2-hydroxyglutarate, a vital product of the isocitrate dehydrogenase mutation in a subset of gliomas. As a technological step in the project proposals, we successfully designed, developed, and implemented two parallel technical projects. Project #1: In the first project, we aimed to develop an innovative sequence program for whole-brain fast magnetic resonance spectroscopic imaging (MRSI) for clinical application specifically, but not limited to, gliomas studies. The improvement was investigated on phantoms containing brain metabolites, healthy volunteers, and a number of glioma patients recruited at MGH. The results were promising, providing up to 40% improvement in the signal-to-noise ratio (SNR) while the sequence program was performed at the same scanning window time. The MR experiments were performed on both clinical 3T MRI and 7T MRI systems. The preliminary results from this project were published as a conference article at an international magnetic resonance in medicine annual conference. The conference article is accepted for an oral presentation this April. Project #2: In this study, we aimed to combine the developed MRSI sequence in the first project with a custom in house 32-channel shim array coil. The coil was designed and engineered at Martinos’ center for biomedical imaging in collaboration with the electrical engineering department at Massachusetts Institute of Technology (MIT), Boston, MA, USA. The major findings from this project were multiple improvements on critical parameters of the MRSI modality, including SNR, the linewidth of the metabolites signals, increased number of usable voxels in the brain region of interest, and reduced signal fitting errors. The project was constructed on a multidisciplinary approach which expertise from different background seeking new applications in using the designed hardware (the 32-channel coil) with our software (the developed MRSI in the project 1 and the developed software for field inhomogeneity correction real-time). The MRSI pulse program with the 32-channel coil was performed on a 7T MRI system. The results from this project were also published in the same conference and is accepted for oral presentation this April. We published two articles in 2019 in the field of cardiovascular metabolomics. The studies demonstrate the use of magnetic resonance spectroscopy in the investigation of heart failure in myocardium bioenergetics and metabolism. The results were published in Metabolites and Scandinavian Cardiovascular Journal.

Following the project plans, the candidate spent most of 2019 (8 months) in the United States. The host was the Radiology Department, Massachusetts General Hospital (MGH), Boston, MA, USA; and Harvard Medical School, Boston, MA, USA. Also, the research lab was located at the Martinos’ center for Biomedical Imaging affiliated by MGH/Harvard, Boston, MA, USA. As a technological step in the project proposals, we successfully designed and implemented two parallel technical projects.

2018
The purpose of the project is to investigate the role of molecular imaging in more accurate diagnosis, prognosis, and response to therapy assessment in cancer. Recent discoveries in molecular-based classifications and novel molecular-targeted therapies have provided new insights into gliomagenesis with a high potential in improving cancer medicine.Magnetic resonance spectroscopic imaging (MRSI) provides non-invasive in vivo metabolic profiling of neurologic pathologies such as tumors. Proton MRSI can detect metabolic information from several metabolic pathways of importance in cancer. Advanced and optimized versions of MRSI enable detection of 2-hydroxyglutarate, a vital product of the isocitrate dehydrogenase mutation in a subset of gliomas. Recent studies have shown a favorable overall survival in this subgroup of patients. The hypothesis is that targeting a specific metabolic pathway will benefit IDH mutated gliomas. Thus, an advanced imaging methodology for in vivo molecular imaging of this pathway will help diagnostic, and monitoring of therapy in patients with IDH mutated gliomas. Such an imaging technique will also enhance the detection of other types of cancer and improve individualized cancer therapies. In 2018, we established an analytical platform to investigate the dominant growth directions of cancer cells in a subgroup of patients with brain tumors. The study employed the most recent advances in the field of image analysis to examine the tumor expansion trend in the central nervous system. The results from this study are published in Nature Scientific Reports journal. We also demonstrated that the MRSI technique could benefit molecular imaging of cancer in the other organs such as the prostate. When we combined the data acquired from MRSI with those of positron emission tomography imaging, the diagnostic performance was significantly improved. The result from this investigation is published in Frontier Oncology journal.

NO

Vitenskapelige artikler
Esmaeili M, Vettukattil R

In Vivo Magnetic Resonance Spectroscopy Methods for Investigating Cardiac Metabolism.

Metabolites 2022 Feb 18;12(2). Epub 2022 feb 18

PMID: 35208262

Esmaeili M, Vettukattil R, Banitalebi H, Krogh NR, Geitung JT

Explainable Artificial Intelligence for Human-Machine Interaction in Brain Tumor Localization.

J Pers Med 2021 Nov 16;11(11). Epub 2021 nov 16

PMID: 34834566

Naik M, Esmaeili M, Thomas O, Geitung JT

Diffusion tension imaging is a good tool for assessing patients with dementia and behavioral problems and discriminating them from other dementia patients.

Acta Radiol Open 2021 Dec;10(12):20584601211066467. Epub 2021 des 17

PMID: 34950511

Esmaeili M, Stockmann J, Straßer B, Arango N, Thapa B, Wang Z, van der Kouwe A, Dietrich J, Cahill DP, Batchelor TT, White J, Adalsteinsson E, Wald L, Andronesi OC

An integrated RF-receive/B

Sci Rep 2020 09 14;10(1):15029. Epub 2020 sep 14

PMID: 32929121

Esmaeili M, Straßer B, Bogner W, Moser P, Wang Z, Andronesi OC

Whole-Slab 3D MR Spectroscopic Imaging of the Human Brain With Spiral-Out-In Sampling at 7T.

J Magn Reson Imaging 2020 Nov 12. Epub 2020 nov 12

PMID: 33179836

Shi M, Ellingsen Ø, Bathen TF, Høydal MA, Stølen T, Esmaeili M

The Effect of Exercise Training on Myocardial and Skeletal Muscle Metabolism by MR Spectroscopy in Rats with Heart Failure.

Metabolites 2019 Mar 19;9(3). Epub 2019 mar 19

PMID: 30893827

Stølen T, Shi M, Wohlwend M, Høydal MA, Bathen TF, Ellingsen Ø, Esmaeili M

Effect of exercise training on cardiac metabolism in rats with heart failure.

Scand Cardiovasc J 2020 Apr;54(2):84-91. Epub 2019 sep 10

PMID: 31500456

Esmaeili M, Stensjøen AL, Berntsen EM, Solheim O, Reinertsen I

The Direction of Tumour Growth in Glioblastoma Patients.

Sci Rep 2018 01 19;8(1):1199. Epub 2018 jan 19

PMID: 29352231

Esmaeili M, Tayari N, Scheenen T, Elschot M, Sandsmark E, Bertilsson H, Heerschap A, Selnæs KM, Bathen TF

Simultaneous

Front Oncol 2018;8():516. Epub 2018 nov 15

PMID: 30498693

van den Elshout R, Ariëns B, Blaauboer J, Meijer FJA, van der Kolk AG, Esmaeili M, Scheenen TWJ, Henssen DJHA

Quantification of perineural satellitosis in pretreatment glioblastoma with structural MRI and a diffusion tensor imaging template

Neuro-Oncology Advances, 2023

Esmaeili M, Salami A

Brain functional fingerprinting predicts individual differences in cognition: An AI-based approach

Joint Annual Meeting ISMRM-ESMRMB, 2022

Hansen T, Esmaeili M, Salami A

Prediction of Behavior and Cognition by an AI-based approach

BME@LiU, 2022

Esmaeili M, Salami A

Connectome-based individualized Prediction of Behavior and Cognition by an AI-based approach

Organization for Human Brain Mapping, 2022

Kim E, Esmaeili M, Moestue SA, Bathen TF

MR-Derived Biomarkers for Cancer Characterization. I: Biomarkers of the Tumor Microenvironment

Springer ISBN 978-3-030-98950-7, 2022

Esmaeili M, Antun V, Vettukattil MR, Banitalebi H, Krogh NR, Geitung JT

Evaluation of Automated Brain Tumor Localization by Explainable Deep Learning Methods

ISMRM & SMRT Virtual Conference & Exhibition, 2021. Abstract # 2435

Kim E, Esmaeili M, Moestue S, Bathen T

Book Chapter: MR-Derived Biomarkers for Cancer Characterization, Book title: Biomarkers of the Tumor Microenvironment: Basic Studies and Practical Applications, edited by Lars A. Akslen and Randolph S. Watnick.

2021

Bjørkeli EB, Geitung JT, Esmaeili M

Domain-Transformation of MRI-derived Time Series with Deep Learning

First Nordic Conference for Young AI researchers, 2021

Esmaeili M, Straßer B, Bogner W, Moser P, Wang Z, and Andronesi OC

Whole-brain MR Spectroscopic Imaging with stack of Spirals Out-In k-space Trajectory at 7T

ISMRM & SMRT Virtual Conference & Exhibition, 2020. Abstract # 0370

Esmaeili M, Stockmann J, Straßer B, Wang Z, Wald L, and Andronesi OC

An Integrated RF-receive/B0-shim Array Coil Improves whole-brain MR Spectroscopic Imaging at 7T

ISMRM & SMRT Virtual Conference & Exhibition, 2020. Abstract #4232

Doktorgrader
Mingshu Shi

MRS-based Metabolic profiling of cardiac and skeletal muscle from rats with heart failure

Disputert:
desember 2019
Hovedveileder:
Tomas Stølen
Deltagere
  • Morteza Esmaeili Prosjektleder
  • Jonn Terje Geitung Forsker (annen finansiering)
  • Muhammad Riyas Vettukattil Forsker (annen finansiering)
  • Morteza Esmaeili Prosjektleder
  • Zhe Wang Forsker (annen finansiering)
  • Tracy Batchelor Forsker (annen finansiering)
  • Bernhard Straßer Forsker (annen finansiering)
  • Jason Stockmann Forsker (annen finansiering)
  • Ovidiu Andronesi Forsker (annen finansiering)
  • Nina Krogh Prosjektdeltaker
  • Ole Solheim Prosjektdeltaker
  • Daniel P Cahill Prosjektdeltaker
  • Elizabeth Gerstner Prosjektdeltaker
  • Ingerid Reinertsen Forsker (annen finansiering)
  • Hasan Banitalebi Forsker (annen finansiering)

eRapport er utarbeidet av Sølvi Lerfald og Reidar Thorstensen, Regionalt kompetansesenter for klinisk forskning, Helse Vest RHF, og videreutvikles av de fire RHF-ene i fellesskap, med støtte fra Helse Vest IKT

Alle henvendelser rettes til eRapport

Personvern  -  Informasjonskapsler