The Evolving Landscape of Neuro-Oncology: A Comprehensive Review of Multi-Parametric Magnetic Resonance Imaging (mpMRI)

Many thanks to our sponsor Esdebe who helped us prepare this research report.

Abstract

Multi-Parametric Magnetic Resonance Imaging (mpMRI) stands as a profound advancement in the diagnostic and prognostic arsenal for central nervous system (CNS) pathologies, most notably brain tumors. By meticulously integrating a suite of advanced MRI sequences—including T1-weighted (T1w) pre- and post-contrast, T2-weighted (T2w), Fluid-Attenuated Inversion Recovery (FLAIR), Diffusion-Weighted Imaging (DWI) with Apparent Diffusion Coefficient (ADC) mapping, and Perfusion-Weighted Imaging (PWI) in its various forms—mpMRI transcends the capabilities of conventional anatomical imaging. This sophisticated approach provides a granular, multi-dimensional assessment of tissue microarchitecture, cellularity, vascular dynamics, and metabolic activity, thereby enabling superior tissue differentiation, precise tumor delineation, and enhanced diagnostic accuracy. This comprehensive report meticulously explores the fundamental physical principles underpinning mpMRI sequences, elucidates its significant advantages over standard MRI protocols, details the specific and synergistic information garnered from each imaging modality, and critically examines its indispensable role in the accurate diagnosis, intricate characterization, treatment planning, and dynamic monitoring of brain tumors within demanding clinical and research contexts. Furthermore, the report delves into current clinical applications, acknowledges prevailing challenges and limitations, and projects future trajectories for this transformative imaging paradigm.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

1. Introduction

Magnetic Resonance Imaging (MRI) has, since its inception, revolutionized medical diagnostics by offering unparalleled non-invasive visualization of soft tissues with exquisite anatomical detail. Rooted in the principles of nuclear magnetic resonance, traditional MRI sequences, such as T1w and T2w, have served as the bedrock for identifying structural abnormalities within the brain and spinal cord. These sequences, while providing invaluable anatomical insights, often fall short in offering the requisite biological specificity needed for a definitive characterization of complex pathologies, particularly brain tumors. The inherent limitations of conventional MRI in distinguishing tumor progression from treatment-related effects (e.g., radiation necrosis or pseudo-progression) or in accurately assessing tumor aggressiveness based solely on morphology underscored the pressing need for more sophisticated imaging biomarkers.

The advent of multi-parametric MRI (mpMRI) represents a paradigm shift in neuro-oncological imaging. Rather than relying on a singular tissue contrast, mpMRI synthesizes data from multiple, distinct MRI sequences, each designed to interrogate different physiological and biophysical properties of tissue. This integrated approach allows for the creation of a comprehensive biological fingerprint of a lesion, moving beyond mere anatomical depiction to functional and metabolic characterization. The power of mpMRI lies in its ability to non-invasively probe tumor microenvironment features such as cellular density, vascularity, capillary permeability, and water diffusivity—features that are intimately linked to tumor biology, aggressiveness, and response to therapy. This report aims to provide an exhaustive overview of mpMRI, elucidating its scientific foundations, its transformative impact on neuro-oncology, the practicalities of its implementation, and the exciting prospects it holds for personalized patient management.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

2. Principles of Multi-Parametric MRI: A Deep Dive into Sequence Mechanisms

mpMRI is founded on the strategic acquisition and synthesis of data from several distinct MRI sequences, each meticulously engineered to highlight specific physical and physiological tissue properties. The core principle involves manipulating the inherent magnetic properties of atomic nuclei, primarily protons within water molecules, which are abundant in biological tissues. When placed in a strong external magnetic field (B0), these protons align. Radiofrequency (RF) pulses are then applied to perturb this alignment, and as protons relax back to their equilibrium state, they emit signals that are detected and translated into images. The rate at which these protons relax is characterized by two primary relaxation times: T1 (longitudinal relaxation) and T2 (transverse relaxation), which vary significantly across different tissues and pathological states. mpMRI leverages these fundamental principles through specialized sequences:

2.1 T1-Weighted Imaging (T1w)

Physics and Contrast: T1w sequences are designed with short repetition times (TR) and short echo times (TE) to predominantly reflect differences in longitudinal (T1) relaxation rates. Tissues with short T1 relaxation times appear bright, while those with long T1 times appear dark. Consequently, fat (very short T1) appears bright, white matter (intermediate T1) is brighter than grey matter (longer T1), and cerebrospinal fluid (CSF, very long T1) appears dark. This contrast provides exquisite anatomical detail and is crucial for delineating tissue boundaries.

Clinical Significance in Neuro-Oncology: In brain tumor imaging, T1w sequences are paramount for assessing gross tumor morphology, size, and precise anatomical location. More importantly, T1w images acquired after the intravenous administration of gadolinium-based contrast agents (GBCAs) are vital. GBCAs shorten the T1 relaxation time of tissues where they accumulate. In the context of brain tumors, disruption of the blood-brain barrier (BBB), a hallmark of high-grade malignancy, allows GBCAs to leak into the extravascular extracellular space, leading to areas of bright enhancement on post-contrast T1w images. This enhancement delineates the ‘enhancing tumor’ component, which often correlates with active tumor burden and neovascularization. However, it is critical to note that enhancement is not always synonymous with viable tumor, as certain non-neoplastic conditions (e.g., infection, demyelination) can also cause BBB disruption. Furthermore, significant portions of aggressive brain tumors, particularly gliomas, often extend beyond the enhancing region, a phenomenon known as ‘non-enhancing tumor’ or ‘infiltrative tumor’, which is typically better visualized on FLAIR or T2w sequences.

2.2 T2-Weighted Imaging (T2w)

Physics and Contrast: T2w sequences employ long TR and long TE values, making them sensitive to differences in transverse (T2) relaxation rates. Tissues with long T2 relaxation times, such as water and fluid-rich structures, appear bright. Conversely, tissues with short T2 relaxation times, like white matter, appear darker. This results in CSF appearing bright, grey matter brighter than white matter, and most pathological processes that involve increased water content (e.g., edema, inflammation, necrosis, and many tumors) appearing hyperintense (bright).

Clinical Significance in Neuro-Oncology: T2w images are highly sensitive for detecting pathological changes, including tumor-associated edema, cystic components, and areas of necrosis. Peritumoral edema, often vasogenic in nature, manifests as high signal intensity on T2w images and helps delineate the extent of the tumor’s influence on surrounding brain parenchyma. While sensitive, T2w sequences are often non-specific; various pathologies can appear bright. They are crucial for assessing the overall lesion burden and for guiding biopsy or surgical planning by revealing mass effect and displacement of normal structures.

2.3 Fluid-Attenuated Inversion Recovery (FLAIR)

Physics and Contrast: FLAIR is a specialized inversion recovery sequence that incorporates a specific inversion time (TI) chosen to nullify the signal from free water, particularly CSF. By suppressing the bright signal from CSF, lesions located adjacent to or within CSF spaces become much more conspicuous. The sequence begins with an inversion pulse that flips the net magnetization of protons. After a precise TI, the longitudinal magnetization of CSF protons crosses zero. At this point, a 90-degree excitation pulse is applied, selectively exciting other tissues while the CSF signal is minimized or abolished. The subsequent signal acquisition then emphasizes T2 contrast in all tissues except CSF.

Clinical Significance in Neuro-Oncology: FLAIR is indispensable for visualizing periventricular and subcortical lesions, identifying leptomeningeal carcinomatosis (tumor cells spreading along the meninges, which would otherwise be obscured by bright CSF on standard T2w), and detecting subtle areas of cortical and subcortical edema or tumor infiltration. For gliomas, the hyperintense signal on FLAIR often extends beyond the post-contrast T1w enhancing core, providing a more accurate representation of the infiltrative tumor component and associated vasogenic edema. This ‘FLAIR abnormality’ can encompass both non-enhancing tumor and peritumoral edema, making it a critical component for defining target volumes in radiotherapy and for monitoring disease progression, especially in non-enhancing tumors.

2.4 Diffusion-Weighted Imaging (DWI) and Apparent Diffusion Coefficient (ADC)

Physics and Contrast: DWI measures the microscopic, random motion (Brownian motion) of water molecules within tissues. The sequence applies strong diffusion-sensitizing gradients that cause signal loss from rapidly diffusing water molecules. Areas where water diffusion is restricted (e.g., due to high cellular density, intact cell membranes, or cytotoxic edema) exhibit less signal loss and appear bright on DWI images. However, DWI images are also T2-weighted, meaning a T2 ‘shine-through’ effect can cause genuinely T2-bright lesions to appear bright on DWI, even without restricted diffusion.

To overcome this, an Apparent Diffusion Coefficient (ADC) map is calculated from DWI images acquired with multiple diffusion weighting (b-values). ADC maps represent a quantitative measure of the magnitude of water diffusion, independent of T2 shine-through. Areas of restricted diffusion will appear dark (low ADC values), while areas of facilitated diffusion (e.g., necrosis, vasogenic edema) will appear bright (high ADC values).

Quantitative Metrics and Clinical Significance in Neuro-Oncology: ADC values are highly valuable in neuro-oncology:
* Tumor Cellularity: High-grade brain tumors (e.g., glioblastoma, primary CNS lymphoma, medulloblastoma) often exhibit low ADC values due to high cellular packing density, which impedes water movement. This contrasts with lower-grade tumors or cysts that typically have higher ADC values. ADC can thus serve as a biomarker for tumor grade and aggressiveness.
* Distinguishing Lesions: DWI/ADC helps differentiate acute ischemia (very low ADC) from chronic lesions, and highly cellular tumors from abscesses (which also have low ADC due to viscous pus but distinct clinical presentations). It is also crucial for differentiating cytotoxic edema (low ADC) from vasogenic edema (high ADC).
* Monitoring Treatment Response: Changes in ADC values can reflect treatment efficacy. Successful chemotherapy or radiotherapy leading to tumor cell death or reduced cellularity may result in an increase in ADC values, indicating a positive response.

2.5 Perfusion-Weighted Imaging (PWI)

Physics and Contrast: PWI provides critical insights into tumor vascularity, microvascular density, and blood-brain barrier integrity. There are two primary techniques:
* Dynamic Susceptibility Contrast (DSC-PWI): This is the most commonly used method. It involves the rapid intravenous injection of a bolus of a gadolinium-based contrast agent (a T2/T2 shortening agent) and fast sequential image acquisition. As the bolus passes through the brain capillaries, it causes a transient drop in signal intensity on T2-weighted images. The magnitude and kinetics of this signal drop are analyzed to derive perfusion parameters.
* Arterial Spin Labeling (ASL): This is a non-invasive technique that uses magnetically labeled arterial blood water as an endogenous tracer. RF pulses are applied in the neck to label arterial spins, which then flow into the brain parenchyma, acting as a contrast agent. This allows for quantification of cerebral blood flow (CBF) without exogenous contrast administration, making it suitable for patients with renal impairment or for repeated scans.

Quantitative Metrics and Clinical Significance in Neuro-Oncology: PWI generates various quantitative or semi-quantitative maps:
* Cerebral Blood Volume (CBV): Often presented as relative CBV (rCBV), this parameter reflects the volume of blood within a given tissue area. High-grade brain tumors, characterized by neovascularization and high microvascular density, typically exhibit elevated rCBV values. This is a powerful biomarker for tumor grade, distinguishing high-grade from low-grade gliomas, and differentiating tumor recurrence from radiation necrosis (where rCBV is typically low or normalized).
* Cerebral Blood Flow (CBF): Measures the volume of blood passing through a given volume of tissue per unit time.
* Mean Transit Time (MTT) and Time to Peak (TTP): These parameters reflect the time taken for blood to traverse the capillary bed. Prolonged MTT and TTP can indicate areas of hypoperfusion or compromised blood flow.
* Vascular Permeability (Ktrans): Derived from dynamic contrast-enhanced (DCE-MRI), a T1-based perfusion technique, Ktrans quantifies the transfer rate of contrast agent from the plasma to the extravascular extracellular space, reflecting BBB permeability. High Ktrans values are indicative of a leaky BBB, common in aggressive tumors.

PWI, particularly rCBV, is a cornerstone for grading gliomas, identifying the most aggressive foci within heterogeneous tumors, and crucially, differentiating true tumor progression/recurrence (typically high rCBV) from radiation necrosis (typically low rCBV or similar to normal white matter), a challenging distinction on conventional imaging.

2.6 Magnetic Resonance Spectroscopy (MRS) – An Adjunct Parametric Modality

While not always explicitly listed as a primary mpMRI sequence in the same vein as T1w/T2w, MRS is often considered an integral part of an advanced multi-parametric assessment in neuro-oncology. MRS provides non-invasive biochemical information by detecting and quantifying specific metabolites within a voxel of interest.

Physics and Contrast: MRS exploits the ‘chemical shift’ phenomenon, where the resonance frequency of a proton varies slightly depending on its local chemical environment. This allows for the detection of distinct spectral peaks corresponding to different molecules. Common metabolites analyzed in brain tumors include:
* N-acetylaspartate (NAA): A marker of neuronal viability and density. Decreased NAA is typically seen in tumors due as neurons are replaced or destroyed.
* Choline (Cho): Involved in cell membrane synthesis and turnover. Elevated Cho levels indicate increased cellular proliferation, a hallmark of malignancy.
* Creatine (Cr): A relatively stable marker of cellular energy metabolism, often used as an internal reference.
* Lactate (Lac): An indicator of anaerobic glycolysis, often seen in necrotic or hypoxic tumor regions.
* Lipids: Present in significant amounts in necrotic tissue, reflecting cell membrane breakdown.

Clinical Significance in Neuro-Oncology: In brain tumors, MRS helps characterize metabolic profiles:
* Tumor Grade: High-grade gliomas typically show elevated Cho/NAA ratios (increased proliferation, decreased neuronal density) and sometimes elevated Cho/Cr ratios. Low-grade tumors may show less dramatic changes. The presence of lactate and lipids suggests necrosis.
* Differentiation: MRS can help distinguish tumors from non-neoplastic lesions (e.g., abscesses often show distinct amino acid peaks). It can also aid in differentiating tumor recurrence (high Cho, low NAA) from radiation necrosis (low Cho, relatively preserved NAA, often elevated lactate and lipids).
* Guiding Biopsy: MRS can identify metabolically active regions within heterogeneous tumors, guiding stereotactic biopsy to ensure sampling of the most aggressive component.

By integrating these diverse sequences, mpMRI constructs a holistic view of the tumor microenvironment, providing a level of detail and specificity unattainable by any single modality.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

3. Advantages Over Standard MRI

While standard anatomical MRI (typically T1w and T2w) has been foundational, mpMRI offers profound advantages, significantly elevating diagnostic and prognostic capabilities in neuro-oncology. These advantages stem from its ability to transcend mere structural visualization and delve into the underlying biological processes.

3.1 Enhanced Tissue Characterization and Specificity

Standard MRI provides excellent anatomical context but often lacks the specificity to definitively characterize the nature of a lesion. For instance, both a high-grade glioma and an abscess can appear as enhancing lesions with surrounding edema on conventional T1w post-contrast and T2w images. mpMRI overcomes this by integrating parameters that reflect distinct tissue properties:
* Cellularity vs. Edema: DWI/ADC allows differentiation of high cellularity (e.g., active tumor, lymphoma, abscess core) which causes restricted diffusion (low ADC) from vasogenic edema or necrosis, which typically exhibit facilitated diffusion (high ADC).
* Vascularity and Permeability: PWI provides quantitative measures of microvascular density (rCBV) and blood-brain barrier integrity (Ktrans). This is crucial for distinguishing highly vascularized, aggressive tumors from less vascularized lesions or non-neoplastic processes like demyelination or radiation necrosis.
* Metabolic Fingerprinting: MRS provides a metabolic profile, differentiating areas of high cell turnover (high Cho) characteristic of malignancy from neuronal loss (low NAA) or necrosis (lactate, lipids).

This multi-faceted approach creates a unique ‘biomarker signature’ for different pathologies, dramatically improving diagnostic specificity where conventional imaging falters.

3.2 Improved Tumor Detection and Delineation

Tumor detection and precise delineation are critical for accurate diagnosis, surgical planning, and radiation therapy. Standard post-contrast T1w imaging often only captures the ‘enhancing core’ of a tumor, particularly in infiltrative gliomas. However, significant tumor burden, especially in glioblastoma, extends beyond this enhancing region. mpMRI addresses this limitation:
* Identifying Non-Enhancing Infiltrative Tumor: FLAIR hyperintensity extending beyond the T1w enhancing core is crucial for delineating the full extent of peritumoral edema and non-enhancing infiltrative tumor cells, which are biologically active and contribute to recurrence. This is paramount for defining the Clinical Target Volume (CTV) in radiotherapy.
* Mapping Heterogeneity: mpMRI allows for the segmentation of various tumor sub-regions—enhancing tumor, non-enhancing tumor, necrosis, and peritumoral edema—each with distinct biological characteristics. This comprehensive mapping aids in precise surgical planning (identifying eloquent areas), guiding targeted biopsies to the most aggressive foci, and ensuring adequate radiation coverage.

3.3 Assessment of Tumor Heterogeneity and Aggressiveness

Brain tumors, especially high-grade gliomas, are notoriously heterogeneous, comprising regions with varying cellularity, vascularity, hypoxia, and genetic profiles. This intratumoral heterogeneity significantly impacts treatment response and patient prognosis. Standard MRI often averages these variations. mpMRI, however, excels at characterizing this heterogeneity:
* Spatial Mapping of Aggressiveness: PWI can pinpoint regions of maximal rCBV, indicating areas of peak microvascular density and potential aggressiveness within a larger tumor. Similarly, DWI can identify regions of minimal ADC, correlating with highest cellularity. These ‘hotspots’ may harbor the most malignant clones.
* Predicting Tumor Grade and Molecular Subtypes: mpMRI parameters have shown correlations with histopathological tumor grade (e.g., WHO classification) and increasingly, with specific molecular markers (e.g., IDH mutation status, 1p/19q co-deletion, MGMT promoter methylation) which are crucial for classification and prognosis. Radiomics, the extraction of high-throughput quantitative features from mpMRI images, further enhances this predictive capability.
* Monitoring Therapeutic Response: By tracking changes in mpMRI parameters (e.g., changes in rCBV or ADC) over time, clinicians can assess the efficacy of therapeutic interventions earlier than overt changes in tumor size on anatomical images. For example, a decrease in rCBV may indicate an anti-angiogenic response, while an increase in ADC might suggest tumor necrosis following chemotherapy.

3.4 Differentiating Treatment-Related Changes from True Progression

One of the most challenging diagnostic dilemmas in neuro-oncology is distinguishing true tumor recurrence or progression from treatment-related changes such as radiation necrosis or pseudo-progression (transient increase in enhancing lesion size following chemoradiation). Both can mimic tumor progression on conventional post-contrast T1w images, leading to potentially inappropriate treatment escalation or cessation.
* Radiation Necrosis vs. Tumor Recurrence: mpMRI is invaluable here. Radiation necrosis typically exhibits low rCBV on PWI and high ADC values on DWI (due to fluid-rich necrotic tissue), whereas tumor recurrence generally shows high rCBV (neovascularization) and low ADC (high cellularity). MRS often reveals low Cho and normal NAA in radiation necrosis, contrasting with high Cho/NAA ratios in recurrent tumor.
* Pseudo-progression: This phenomenon, particularly common after temozolomide and radiation for glioblastoma, involves an initial increase in enhancement and edema that stabilizes or improves spontaneously without true tumor growth. mpMRI can help differentiate this benign entity from true progression by demonstrating stable or decreasing rCBV and ADC values within the enhancing lesion, or even showing reduced vascularity compared to a true recurrence.

In essence, mpMRI provides a deeper, more specific, and dynamic understanding of tumor biology, transforming the diagnostic and management pathways for brain tumor patients.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

4. Information Captured by mpMRI Sequences: Detailed Contributions

Each component of the mpMRI protocol contributes unique, synergistic information, creating a comprehensive picture of the underlying pathology:

4.1 T1w Imaging: Anatomical Blueprint and Barrier Integrity

As previously detailed, pre-contrast T1w images provide the anatomical framework, delineating brain structures, skull, and basic lesion morphology. Post-contrast T1w imaging, however, provides the critical information regarding blood-brain barrier (BBB) integrity. Enhancing regions signify BBB disruption, indicative of active tumor growth, inflammation, or infection. The volume and pattern of enhancement are often correlated with tumor grade, with irregular, thick, and rim-enhancing patterns typically seen in high-grade gliomas and metastases. This sequence is fundamental for initial diagnosis, assessing mass effect, and tracking the gross enhancing component of the tumor over time.

4.2 T2w Imaging: Edema, Cysts, and Gross Pathology

T2w sequences are highly sensitive to tissue water content. In neuro-oncology, they primarily highlight peritumoral edema (vasogenic edema due to BBB breakdown around the tumor) and cystic/necrotic components within the tumor. High signal intensity on T2w images delineates the overall lesion extent and its impact on surrounding brain parenchyma, including mass effect and hydrocephalus. While sensitive, T2w is non-specific; therefore, it requires correlation with other sequences for definitive characterization. It is particularly useful for identifying non-enhancing tumor components that have high water content but no BBB disruption.

4.3 FLAIR Imaging: CSF Nulling and Peritumoral Infiltration

FLAIR sequences are invaluable for suppressing the high signal from CSF, which can otherwise obscure lesions in the periventricular regions or subarachnoid space. This sequence is particularly adept at visualizing:
* Peritumoral Edema and Infiltrative Tumor: For gliomas, the hyperintensity on FLAIR often extends significantly beyond the enhancing margin on post-contrast T1w, representing a combination of vasogenic edema and infiltrative non-enhancing tumor cells. This ‘FLAIR lesion’ provides a more accurate representation of the total tumor burden for treatment planning.
* Leptomeningeal Carcinomatosis: Subtle tumor deposits along the meninges or within the CSF spaces, which would be challenging to detect on standard T2w due to bright CSF, become clearly visible as hyperintense lesions on FLAIR.
* Non-Enhancing Tumors: In lower-grade gliomas that do not enhance after contrast, FLAIR is often the primary sequence for delineating tumor extent.

4.4 DWI and ADC Mapping: Cellularity and Microstructural Integrity

DWI and its quantitative derivative, ADC, are powerful probes of the cellular microenvironment:
* Restricted Diffusion (Low ADC): Indicates impediments to water molecule movement. This is typically observed in:
* High Cellularity: Dense packing of cells, as seen in high-grade gliomas (e.g., glioblastoma, anaplastic astrocytoma, primary CNS lymphoma, medulloblastoma), germinomas, and meningiomas. The small extracellular space restricts water movement.
* Cytotoxic Edema: As seen in acute ischemia, where cell swelling reduces the extracellular space.
* Abscesses: Due to the high viscosity of pus.
* Facilitated Diffusion (High ADC): Suggests increased extracellular space and less impedance to water movement, often seen in:
* Vasogenic Edema: Accumulation of fluid in the extracellular space due to BBB disruption.
* Necrosis/Cysts: Fluid-filled cavities with minimal cellularity.
* Low-grade Tumors: Often less cellular than high-grade counterparts.

ADC values are quantitatively useful for predicting tumor grade (lower ADC often correlates with higher grade) and distinguishing different types of lesions. The spatial distribution of ADC values can highlight regions of highest cellularity within a heterogeneous tumor, which can be targeted for biopsy or aggressive therapy.

4.5 PWI: Hemodynamics and Angiogenesis

PWI measures regional hemodynamics, providing insights into the tumor’s vascular supply and its capacity for angiogenesis. Key parameters derived are:
* Relative Cerebral Blood Volume (rCBV): This is arguably the most valuable PWI parameter in neuro-oncology. High rCBV indicates increased microvascular density and neovascularization, a hallmark of aggressive, high-grade tumors. It is a robust predictor of tumor grade and a critical tool for differentiating tumor recurrence from radiation necrosis (recurrent tumor has high rCBV, necrosis has low rCBV).
* Relative Cerebral Blood Flow (rCBF): Reflects the rate of blood delivery. While correlated with rCBV, rCBV often provides a stronger differentiation for tumor grading due to its sensitivity to capillary density rather than just flow.
* Mean Transit Time (MTT) and Time to Peak (TTP): Prolonged MTT and TTP can indicate areas of compromised perfusion or tortuous vessels within tumors.
* Ktrans (from DCE-MRI): Measures the transfer rate of contrast agent across the leaky BBB into the extravascular space. High Ktrans indicates increased vascular permeability, a feature of aggressive tumors and a potential marker of response to anti-angiogenic therapies.

By integrating these hemodynamic parameters, PWI offers a functional assessment of tumor angiogenesis, which is vital for understanding tumor biology, predicting its aggressiveness, and monitoring the effectiveness of therapies targeting the tumor vasculature.

4.6 MRS: Metabolic Profile

As an adjunct, MRS adds a metabolic dimension. The ratios of key metabolites like Cho/NAA and Cho/Cr provide a biochemical fingerprint:
* Increased Cho/NAA & Cho/Cr: Highly indicative of increased cell membrane synthesis (proliferation) and neuronal destruction, characteristic of high-grade malignancy.
* Lactate and Lipids: Often present in necrotic or hypoxic tumor regions, reflecting anaerobic metabolism and cell death.

MRS can help differentiate tumor types, grade tumors, and distinguish viable tumor from treatment effects, complementing the anatomical and physiological information from other mpMRI sequences.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

5. Role in Brain Tumor Characterization: A Comprehensive Approach

mpMRI has become an indispensable tool throughout the entire patient journey for brain tumor management, from initial diagnosis to long-term surveillance. Its multi-parametric nature allows for a refined understanding of tumor biology that informs critical clinical decisions.

5.1 Accurate Diagnosis and Differential Diagnosis

Beyond simply detecting a lesion, mpMRI significantly aids in the differential diagnosis of intracranial masses. By combining morphological features from T1w, T2w, and FLAIR with microstructural information from DWI (cellularity), hemodynamic data from PWI (vascularity), and metabolic profiles from MRS, mpMRI can help distinguish:
* Primary Brain Tumors (e.g., Gliomas, Lymphoma): Different grades of gliomas often have distinct mpMRI signatures. For instance, high-grade gliomas typically show marked enhancement, central necrosis, prominent peritumoral FLAIR hyperintensity, low ADC in solid tumor components, and elevated rCBV. Primary CNS lymphoma also shows restricted diffusion (low ADC) due to extreme cellularity but often has less prominent neovascularization than glioblastoma.
* Metastatic Lesions: Often appear as well-defined, spherical enhancing lesions, sometimes with surrounding vasogenic edema. PWI may show high rCBV, similar to high-grade gliomas.
* Non-Neoplastic Lesions:
* Abscesses: Typically show ring enhancement on T1w post-contrast, bright signal on T2w/FLAIR, but critically, very low ADC (restricted diffusion) within the central cavity due to viscous pus, which differentiates them from necrotic tumors.
* Demyelinating Lesions (e.g., Multiple Sclerosis): May show enhancement but typically lack significant mass effect, have specific FLAIR signal characteristics, and generally do not demonstrate the high rCBV or very low ADC seen in aggressive tumors.
* Infarcts: Acute infarcts show profound restricted diffusion (very low ADC), and characteristic vascular territory distribution.

5.2 Tumor Grading and Prognostication

mpMRI parameters provide surrogate biomarkers for tumor grade and can correlate with molecular markers, offering prognostic insights:
* Correlation with WHO Grading: Quantitative parameters such as minimum ADC values (inversely correlated with cellularity) and maximum rCBV values (directly correlated with microvascular density) are highly predictive of WHO histological grade, particularly for gliomas. Lower ADC and higher rCBV generally correspond to higher grades.
* Prediction of Molecular Markers: Radiogenomics, an emerging field, utilizes advanced image analysis (radiomics) of mpMRI data to predict the presence of specific genetic mutations or biomarkers (e.g., IDH mutation status, 1p/19q co-deletion, MGMT promoter methylation) that are crucial for tumor classification, prognosis, and treatment selection. For instance, IDH-mutant gliomas often exhibit a more infiltrative pattern, less enhancement, and higher ADC values compared to IDH-wildtype counterparts.
* Survival Prediction: Specific mpMRI features or derived radiomic signatures have been shown to correlate with overall survival and progression-free survival in various brain tumor types, aiding in personalized prognostication.

5.3 Treatment Planning and Biopsy Guidance

Accurate tumor delineation is paramount for neurosurgical resection and radiation therapy planning. mpMRI provides the necessary spatial and biological information:
* Surgical Planning: mpMRI allows neurosurgeons to meticulously delineate the enhancing tumor core, peritumoral edema, and non-enhancing infiltrative regions. Functional MRI (fMRI) can be integrated to map eloquent cortical areas (e.g., motor, language cortex) relative to the tumor, facilitating maximal safe resection while preserving neurological function. DWI-derived Diffusion Tensor Imaging (DTI) can map white matter tracts, informing approaches to minimize neurological deficits.
* Radiation Therapy Planning: For radiation oncologists, mpMRI provides critical input for defining Gross Tumor Volume (GTV), Clinical Target Volume (CTV), and Planning Target Volume (PTV). The FLAIR abnormality often dictates the CTV for gliomas, ensuring that infiltrative components beyond the enhancing core are adequately irradiated. PWI and DWI can identify biologically aggressive sub-regions to be targeted with higher radiation doses (dose painting).
* Targeted Biopsy: In heterogeneous tumors, mpMRI (especially PWI and DWI) can guide stereotactic biopsies to the most aggressive or metabolically active tumor regions, increasing the yield of diagnostic tissue and ensuring representation of the highest-grade component.

5.4 Monitoring Treatment Response and Recurrence

One of the most impactful roles of mpMRI is in the post-treatment surveillance of brain tumors, particularly in distinguishing true tumor progression/recurrence from treatment-related changes. This distinction is vital as the management strategies are diametrically opposed.
* Pseudo-progression vs. True Progression: After chemoradiation for glioblastoma, a transient increase in tumor enhancement and edema (pseudo-progression) can occur. mpMRI, specifically PWI and DWI, helps differentiate this from true tumor progression. Pseudo-progression typically shows stable or decreased rCBV and often increased ADC within the enhancing lesion, whereas true progression shows increasing rCBV and stable or decreasing ADC.
* Radiation Necrosis vs. Tumor Recurrence: As discussed, radiation necrosis typically presents with low rCBV and high ADC, contrasting with the high rCBV and low ADC characteristic of recurrent tumor. MRS profiles further aid this differentiation (low Cho for necrosis, high Cho for recurrence).
* Assessing Response to Novel Therapies: mpMRI provides quantitative biomarkers to assess the effectiveness of new drugs, especially anti-angiogenic agents. A decrease in rCBV or Ktrans, even without changes in tumor size, can indicate a positive anti-angiogenic response. Changes in ADC can reflect cytotoxic effects. These early markers can guide decisions on continuing or altering therapy.
* Surveillance: For long-term follow-up, mpMRI provides a sensitive means to detect subtle changes indicative of recurrence, allowing for timely intervention.

By offering a nuanced and dynamic view of tumor biology, mpMRI empowers clinicians to make more informed, personalized decisions throughout the continuum of brain tumor care.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

6. Clinical Applications and Case Studies: Empirical Evidence

The utility of mpMRI in neuro-oncology is extensively supported by a growing body of scientific literature and its integration into clinical guidelines. Its application spans across various facets of tumor management, with significant advancements in automated analysis.

6.1 Advancements in Automated Segmentation

Accurate and reproducible segmentation of brain tumors and their sub-regions from mpMRI data is critical for diagnosis, treatment planning, and monitoring. Traditionally, this was a laborious and subjective manual process. mpMRI, with its rich feature set, has paved the way for advanced computational methods:

• Semi-Automated Segmentation with Non-Negative Matrix Factorization (NMF): A notable study demonstrated the efficacy of a semi-automated framework utilizing Non-Negative Matrix Factorization (NMF) on mpMRI data for brain tumor segmentation. NMF is a dimensionality reduction technique that can decompose a multi-parametric dataset into a set of ‘basis components’ and ‘mixing coefficients’, which can then be interpreted as representing distinct tissue types or biological features within the tumor. In this specific study, NMF was applied to T1w, T1w post-contrast, T2w, and FLAIR sequences. The framework achieved mean Dice scores of 65% for active tumor, 74% for the tumor core (enhancing tumor + necrosis), and 80% for the whole tumor region (enhancing tumor + necrosis + non-enhancing/edema) [1]. Dice score, a measure of overlap between automated and ground truth segmentations, indicates good agreement. This semi-automated approach significantly reduced the subjectivity and time associated with manual segmentation, highlighting mpMRI’s potential to enhance segmentation accuracy by leveraging the multi-modal information. The clinical implication is more consistent and reliable tumor volume measurements, leading to better-informed treatment planning.

• Automated Tumor Segmentation in Pediatrics using Deep Learning: Pediatric brain tumors present unique challenges due to diverse pathologies, varying anatomical structures, and the need for sedation during imaging. A multi-institutional study addressed these challenges by applying deep learning-based methods to mpMRI scans of pediatric brain tumors [2]. The study utilized a diverse dataset of mpMRI sequences (T1w pre- and post-contrast, T2w, FLAIR) from multiple institutions, allowing for robust model training. Deep learning architectures, such as sophisticated Convolutional Neural Networks (CNNs) or U-Net variants, were employed to learn intricate patterns from the multi-parametric data. The models achieved impressive median Dice similarity scores of 0.91 for whole tumor segmentation. This demonstrates the critical utility of mpMRI in conjunction with AI for overcoming the complexities of pediatric neuro-oncology, enabling precise delineation for surgical planning and radiation therapy in a vulnerable patient population. The high Dice score indicates excellent agreement between automated and expert manual segmentations, showcasing the reliability of these AI-driven mpMRI analyses.

• Confidence-Guided Enhancing Brain Tumor Segmentation: Another study focused on improving the robustness of enhancing brain tumor segmentation by incorporating a ‘confidence-guided’ approach [8]. Enhancing tumor regions, though often small, are crucial for assessing treatment response and identifying areas of active disease. Deep learning models can sometimes exhibit overconfidence or errors in these critical areas. The confidence-guided framework likely involves estimating the uncertainty associated with each segmented voxel, allowing the model to flag areas where its segmentation is less certain. By leveraging mpMRI sequences (e.g., post-contrast T1w, FLAIR), this approach aimed to improve the accuracy and reliability of segmenting enhancing tumor components, particularly in challenging cases, thereby enhancing the trustworthiness of automated segmentation outputs in clinical decision-making.

6.2 Broader Clinical Utility

Beyond segmentation, mpMRI has transformed various aspects of brain tumor management:
* Pre-surgical Mapping: Integration of mpMRI with functional MRI (fMRI) for eloquent cortex mapping and Diffusion Tensor Imaging (DTI) for white matter tractography allows neurosurgeons to meticulously plan resections to achieve maximal tumor removal while preserving critical neurological functions. This personalized approach is a direct outcome of mpMRI’s ability to provide both structural and functional insights.
* Monitoring Response to Anti-Angiogenic Therapies: In patients receiving anti-angiogenic agents (e.g., bevacizumab) for glioblastoma, conventional imaging may show a reduction in enhancement due to decreased vascular permeability, which could be misinterpreted as tumor shrinkage. However, mpMRI with PWI can differentiate this ‘pseudo-response’ from true tumor response. A reduction in rCBV on PWI would indicate a true anti-angiogenic effect, whereas stable or increasing rCBV despite reduced enhancement would suggest continued tumor viability and infiltration, necessitating a change in treatment strategy.
* Differentiation of Recurrence vs. Radionecrosis: As highlighted, this is a major clinical challenge. A meta-analysis examining the diagnostic accuracy of various mpMRI parameters (PWI-derived rCBV, DWI-derived ADC, and MRS-derived Cho/NAA ratios) showed that these combined parameters significantly improve the ability to distinguish tumor recurrence from radiation necrosis, with PWI often demonstrating the highest individual diagnostic accuracy. This informs critical decisions on re-operation, re-irradiation, or conservative management.

These examples underscore how mpMRI, by providing a wealth of quantitative and qualitative data, empowers clinicians to make more accurate diagnoses, devise more effective treatment strategies, and dynamically monitor disease behavior, ultimately leading to improved patient outcomes.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

7. Challenges and Limitations

Despite its transformative potential, mpMRI is not without its challenges and limitations that warrant careful consideration in both clinical practice and research.

7.1 Technical Challenges and Standardization

• Inter-Scanner and Inter-Institutional Variability: One of the most significant challenges is the lack of universal standardization in mpMRI acquisition protocols. Different MRI scanner vendors, field strengths (1.5T, 3T, 7T), pulse sequence parameters (e.g., TR, TE, TI, b-values for DWI, flip angles, contrast injection rates for PWI), and coil configurations can lead to substantial variability in image quality, contrast, and quantitative metrics (e.g., ADC values, rCBV values). This variability makes it difficult to compare studies across different centers, hampers the development of generalizable AI models, and complicates multi-center clinical trials.
• Motion Artifacts: The acquisition of multiple sequences can be lengthy, increasing the likelihood of patient motion, which can degrade image quality and lead to misregistration between different parametric maps, hindering accurate co-registration and analysis.
• Susceptibility Artifacts: Certain mpMRI sequences, particularly T2* and PWI, are sensitive to susceptibility artifacts caused by metallic implants (e.g., surgical clips, dental fillings) or hemorrhage, which can obscure critical anatomical areas.
• Contrast Agent Kinetics and Leakage Correction: DSC-PWI requires careful consideration of contrast agent bolus timing and dosage. Furthermore, contrast leakage from a disrupted BBB into the extravascular space can artificially increase the T1 signal and confound T2*-based CBV calculations, leading to overestimation of rCBV. Advanced post-processing techniques are needed to correct for this leakage effect, but these methods can be complex and are not universally implemented.
• Longer Acquisition Times: The need to acquire multiple high-resolution sequences inherently increases the overall scan duration. This can be problematic for patients who are critically ill, claustrophobic, or unable to remain still for extended periods, potentially requiring sedation and reducing patient throughput.

7.2 Data Interpretation Complexity and Expertise

• Data Overload and Integration: mpMRI generates a vast amount of complex imaging data. Integrating and interpreting information from multiple parametric maps (e.g., T1w post-contrast, FLAIR, ADC, rCBV, MRS spectra) requires specialized expertise, advanced training, and a thorough understanding of neuro-oncology. The interpretation is often more qualitative than quantitative in routine clinical practice, leading to inter-reader variability.
• Post-Processing Variability: Different commercial or open-source software packages for generating parametric maps (e.g., ADC, rCBV) can employ varying algorithms and normalization strategies, leading to discrepancies in quantitative values. This lack of standardization in post-processing further compounds the interpretation challenges.
• Lack of Universal Quantitative Biomarkers: While promising, many mpMRI parameters are still considered ‘relative’ or semi-quantitative. Definitive, universally accepted cut-off values for distinguishing pathologies or predicting outcomes are often lacking or vary with acquisition parameters, limiting their absolute diagnostic certainty in all cases.

7.3 Accessibility, Cost, and Resource Intensiveness

• Cost and Availability: mpMRI protocols are typically more expensive and require more specialized equipment and technical support than standard MRI. This limits their widespread availability, particularly in resource-constrained settings.
• Computational Demands: The advanced post-processing and analysis techniques, especially those involving AI/ML, require significant computational power and expertise, which may not be readily available in all clinical environments.

7.4 Specific Challenges for Pediatric Populations

• Anesthesia Requirements: Younger pediatric patients often require sedation or general anesthesia for mpMRI, adding risks and logistical complexities.
• Developmental Changes: The developing brain exhibits dynamic changes in water content, myelination, and metabolism, which can influence mpMRI parameters and make interpretation more complex compared to adult brains.
• Tumor Heterogeneity in Pediatrics: Pediatric brain tumors are highly diverse, often requiring different mpMRI interpretation strategies for various tumor types.

Addressing these limitations through standardization efforts, development of robust and user-friendly analysis tools, and continued research into quantitative biomarkers will be crucial for maximizing the clinical impact of mpMRI.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

8. Future Directions: Pioneering the Next Generation of Neuro-Oncology Imaging

The field of mpMRI is in a continuous state of evolution, with exciting advancements on the horizon poised to further enhance its capabilities and impact on neuro-oncology.

8.1 Advanced Image Acquisition and Reconstruction Techniques

• Ultra-High Field MRI (7T and Beyond): Higher magnetic field strengths (7 Tesla and up) offer increased signal-to-noise ratio (SNR) and enhanced spatial resolution, enabling the visualization of finer anatomical details and more precise quantification of metabolic and physiological parameters. This could lead to earlier detection of subtle lesions and more granular insights into tumor microarchitecture.
• Accelerated Acquisition Techniques: Techniques such as compressed sensing, parallel imaging (e.g., GRAPPA, SENSE), and simultaneous multi-slice (SMS) imaging are being developed to significantly reduce mpMRI scan times without compromising image quality. Faster acquisitions improve patient comfort, reduce motion artifacts, and increase scanner throughput, making mpMRI more practical for routine clinical use.
• Quantitative MRI (qMRI): The shift from relative or semi-quantitative measurements to absolute quantitative mapping of tissue properties (e.g., quantitative T1, T2, T2* mapping, absolute CBF/CBV, quantitative diffusion parameters) is a major focus. qMRI aims to provide reproducible, scanner-independent biomarkers, facilitating multi-center studies and improving diagnostic consistency.

8.2 Integration with Artificial Intelligence and Machine Learning

The exponential growth in AI and ML, particularly deep learning, is revolutionizing mpMRI analysis:
* Automated Segmentation and Classification: Deep learning algorithms (e.g., U-Net, V-Net, Transformer-based architectures) are increasingly capable of automatically segmenting tumor sub-regions (enhancing tumor, necrosis, edema, non-enhancing tumor) with high accuracy and efficiency, often surpassing human capabilities in consistency. This frees up radiologists’ time and provides objective, reproducible measurements for treatment planning and monitoring.
* Radiomics and Radiogenomics: This involves extracting a vast number of quantitative features (e.g., shape, texture, intensity-based) from mpMRI images. Machine learning models can then correlate these radiomic features with tumor histology, genetic mutations (radiogenomics, e.g., IDH mutation, MGMT methylation), treatment response, and patient survival. This can enable ‘virtual biopsies,’ potentially reducing the need for invasive procedures and facilitating personalized treatment selection.
* Explainable AI (XAI): As AI models become more complex (‘black boxes’), XAI techniques are being developed to provide transparency into how these models arrive at their predictions. This will build trust among clinicians and facilitate the clinical adoption of AI-powered mpMRI tools.
* Federated Learning: This decentralized AI training approach allows multiple institutions to collaboratively train models using local mpMRI datasets without sharing raw patient data, addressing critical privacy and data governance concerns while leveraging large, diverse datasets for robust model development.

8.3 Multi-Modal Integration and Molecular Imaging

• Integration with PET: Combining mpMRI with Positron Emission Tomography (PET), particularly with amino acid tracers (e.g., 18F-FET PET, 11C-MET PET), provides complementary functional information. While mpMRI excels at microstructural and hemodynamic detail, PET offers insights into tumor metabolism and proliferation that mpMRI cannot directly capture. Hybrid PET/MRI scanners enable simultaneous acquisition, ensuring perfect co-registration.
• Integration with Functional MRI (fMRI) and DTI: Beyond basic anatomical localization, fMRI maps brain activity and DTI visualizes white matter tracts. Their seamless integration with mpMRI sequences provides comprehensive functional and structural information crucial for pre-surgical planning, minimizing neurological deficits during tumor resection.
• Targeted Molecular Imaging: Research is ongoing to develop novel MRI contrast agents that specifically target molecular markers expressed by tumor cells (e.g., receptors, enzymes). These ‘smart’ contrast agents could provide highly specific information about tumor biology at a molecular level, allowing for earlier and more precise diagnosis, and potentially guiding targeted drug delivery or theranostics.

8.4 Personalized Medicine and Clinical Decision Support

• Predictive Biomarkers: Future mpMRI applications will focus on developing highly robust predictive biomarkers for response to specific therapies (e.g., chemotherapy, immunotherapy, targeted agents) and for predicting adverse events. This will allow for true personalized medicine, tailoring treatment strategies to individual tumor characteristics.
• Adaptive Radiotherapy: mpMRI, particularly with quantitative parameters, can monitor tumor response and changes in tumor volume/biology during the course of radiotherapy. This real-time information can be used to adapt radiation plans dynamically, optimizing dose delivery and minimizing toxicity to healthy tissue.
• Clinical Decision Support Systems: Integrating AI-powered mpMRI analysis into clinical decision support systems can provide radiologists and oncologists with objective, quantitative data and predictive insights, aiding in complex diagnostic and therapeutic decisions.

The future of mpMRI in neuro-oncology envisions a highly automated, quantitative, and integrated imaging platform that provides comprehensive, actionable insights into tumor biology, ultimately paving the way for more precise, effective, and personalized patient care.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

9. Conclusion

Multi-parametric Magnetic Resonance Imaging (mpMRI) has undeniably emerged as a transformative force in the landscape of neuro-oncology. By ingeniously combining disparate MRI sequences—each a window into distinct facets of tissue biology, from cellularity and microstructural integrity to vascular dynamics and metabolic profiles—mpMRI transcends the limitations of conventional anatomical imaging. It furnishes clinicians with an unprecedented, multi-dimensional understanding of brain tumor characteristics, empowering them to move beyond mere lesion detection to precise diagnosis, nuanced tumor grading, meticulous treatment planning, and dynamic monitoring of therapeutic response.

Its ability to accurately delineate tumor boundaries, assess intratumoral heterogeneity, predict tumor aggressiveness, correlate with molecular biomarkers, and crucially, differentiate true tumor progression from treatment-related changes (such as pseudo-progression or radiation necrosis), represents a monumental leap forward in patient care. The empirical evidence from numerous studies, including the successful implementation of semi-automated and deep learning-based segmentation frameworks, underscores its clinical efficacy across diverse patient populations, including pediatrics.

While challenges persist, particularly concerning standardization of acquisition protocols, the complexity of data interpretation, and computational demands, the trajectory of mpMRI is one of relentless innovation. Future directions, driven by advancements in ultra-high field MRI, accelerated acquisition techniques, quantitative imaging, and the transformative power of Artificial Intelligence and Machine Learning (including radiomics and radiogenomics), promise to further refine its diagnostic and prognostic capabilities. The integration of mpMRI with molecular imaging and other modalities, coupled with the development of sophisticated clinical decision support systems, is poised to usher in an era of truly personalized neuro-oncology. Ultimately, mpMRI stands as a testament to the synergistic interplay of physics, engineering, computer science, and clinical medicine, continually redefining the frontiers of brain tumor diagnosis and management and enhancing the quality of life for countless patients.

Many thanks to our sponsor Esdebe who helped us prepare this research report.

References

1. Menze, B.H., et al. (2010). ‘Semi-automated brain tumor segmentation with non-negative matrix factorization of multi-parametric MRI data.’ Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2010. pubmed.ncbi.nlm.nih.gov
2. Moeskops, P., et al. (2023). ‘Automated brain tumor segmentation in pediatric multi-parametric MRI: A multi-institutional study.’ Neuro-Oncology. pubmed.ncbi.nlm.nih.gov
3. Ellingson, B.M., et al. (2018). ‘Improved assessment of tumor response to antiangiogenic therapy in glioblastoma using dynamic susceptibility contrast perfusion MRI.’ Journal of Clinical Oncology, 36(15), e13568.
4. Hutterer, M., et al. (2020). ‘Predicting IDH mutation and 1p/19q co-deletion in gliomas using deep learning on multiparametric MRI: A multi-institutional study.’ Neuro-Oncology, 22(12), 1774-1784.
5. Gong, C., et al. (2024). ‘Multi-parametric MRI in brain tumor diagnosis and grading.’ Nature Scientific Reports. nature.com
6. Chang, S.M., et al. (2022). ‘RANO criteria for high-grade glioma: Update and recommendations.’ Neuro-Oncology, 24(Supplement_7), vii188-vii194. academic.oup.com
7. Brandão, L.A., et al. (2016). ‘Diffusion-weighted imaging in brain tumors: Current and future applications.’ Neuroradiology, 58(12), 1181-1191. pubmed.ncbi.nlm.nih.gov
8. Kainz, M.S., et al. (2018). ‘Confidence-guided enhancing brain tumor segmentation in multi-parametric MRI using deep learning.’ Medical Image Analysis, 50, 11-23. pure.johnshopkins.edu
9. Louis, D.N., et al. (2021). ‘The 2021 WHO Classification of Tumours of the Central Nervous System: a review.’ Neuro-Oncology, 23(8), 1231-1251.
10. Multispectral segmentation. Wikipedia. en.wikipedia.org
11. Pope, W.B., et al. (2005). ‘Perfusion MR imaging in brain tumors.’ Magnetic Resonance Imaging Clinics of North America, 13(4), 587-601.
12. Law, M., et al. (2008). ‘Glioma grading: sensitivity, specificity, and predictive values of perfusion MR imaging and proton MR spectroscopic imaging for distinguishing low-grade from high-grade gliomas.’ American Journal of Neuroradiology, 29(10), 1888-1896.
13. Kono, K., et al. (2001). ‘The role of diffusion-weighted imaging in patients with brain tumors.’ American Journal of Neuroradiology, 22(6), 1081-1088.
14. Nelson, S.J. (2011). ‘Magnetic resonance spectroscopy and spectroscopic imaging in neuro-oncology.’ Neuroimaging Clinics of North America, 21(1), 77-104.
15. Capper, D., et al. (2018). ‘DNA methylation-based classification of central nervous system tumours.’ Nature, 555(7697), 469-474.
16. Lambin, P., et al. (2012). ‘Radiomics: extracting more information from medical images using advanced feature analysis.’ European Journal of Cancer, 48(4), 441-446.
17. Wen, P.Y., et al. (2010). ‘Updated response assessment criteria for high-grade gliomas: response assessment in neuro-oncology (RANO) working group.’ Journal of Clinical Oncology, 28(11), 1963-1972.
18. Hodi, F.S., et al. (2018). ‘Immunologic and clinical effects of nivolumab in patients with recurrent glioblastoma: results from a phase I clinical trial.’ Clinical Cancer Research, 24(16), 3907-3915.
19. pubmed.ncbi.nlm.nih.gov
20. pure.johnshopkins.edu