Advanced MRI technology
RIBI is challenging to detect early due to its insidious onset and non-specific clinical symptoms. Moreover, distinguishing its imaging features from tumor recurrence remains difficult using a single modality. In recent years, multimodal imaging strategies centered on MRI, integrating structural, functional, metabolic, perfusion, and artificial intelligence analyses, have emerged as crucial approaches for the precise assessment of RIBI (Table 1).
Table 1Comparison of multimodal imaging technologies
Multimodal imaging | Principle | Applications | Pros and Cons |
---|
FLAIR | Suppressing cerebrospinal fluid (CSF) signal to highlight abnormal tissue | Detecting parenchymal abnormalities | Highly sensitive to white matter demyelination |
DWI | Measuring water molecular diffusion | Differentiating hypercellular tumors, radiation necrosis | Provides apparent diffusion coefficient (ADC) values |
DTI | Analyzing the diffusion of water molecules along axonal fibers | Assessing white matter microstructure | Useful for detecting demyelination and axonal loss; Limitation in crossing fiber evaluation |
PWI | Evaluating cerebral perfusion | Monitoring cerebral perfusion | Need for contrast agents: Sensitivity to artifacts. Contrast agents; Longer scan time |
MRS | Quantifying brain metabolite concentration | Quantifying radiation-induced neurotoxicity | Provides non-invasive functional imaging |
BOLD | Based on blood oxygenation level-dependent | Assessing brain function and hemodynamic | Provides non-invasive functional imaging |
Within MRI sequences, T1-weighted imaging, T2-weighted imaging, and fluid-attenuated inversion recovery (FLAIR) are highly sensitive in detecting parenchymal abnormalities, white matter lesions, necrotic foci, and brain atrophy. FLAIR is particularly sensitive to white matter demyelination, often presenting as hyperintense areas within or at the margins of the irradiated field.33
Diffusion imaging techniques reflect the diffusion state of water molecules within tissues. Diffusion-weighted imaging (DWI) can evaluate cellular density and regions with restricted water diffusion, aiding in differentiating hypercellular tumor areas from necrotic tissue. DWI provides apparent diffusion coefficient (ADC) values, also referred to as mean diffusivity. Rapid tumor cell proliferation increases cellular density in lesions, leading to decreased ADC values. In contrast, neuronal necrosis following RIBI reduces cellular density, resulting in increased ADC values.34 Diffusion imaging enables the assessment of microstructural changes in brain tissue. Integration of relative cerebral blood volume (CBV) and Ktrans has achieved diagnostic accuracy exceeding 90% in glioblastoma follow-up.35
Diffusion tensor imaging offers additional parameters, including fractional anisotropy,36,37 axial diffusivity, and radial diffusivity. Several studies have focused on microstructural changes in white matter following radiotherapy, showing decreased fractional anisotropy, increased mean diffusivity,38,39 and elevated radial diffusivity,38,40,41 primarily attributed to demyelination or axonal loss.42,43 Perfusion imaging techniques, by measuring cerebral blood flow and CBV, help differentiate pathological conditions. Studies have shown that regions of radiation necrosis often exhibit hypoperfusion (low cerebral blood flow/CBV), whereas tumor recurrence typically shows hyperperfusion. Among these techniques, arterial spin labeling (ASL) does not require contrast agents and is suitable for repeated monitoring, while dynamic contrast-enhanced (DCE) and dynamic susceptibility contrast (DSC) perfusion imaging provide vascular permeability and perfusion curves, offering high sensitivity in assessing post-radiotherapy lesion activity. Perfusion-weighted imaging (PWI) can detect radiation-induced vascular changes; most PWI studies report reduced perfusion and decreased CBV following radiotherapy.44,45 A combination of PWI and DWI improves diagnostic performance and can obviate unnecessary biopsies in approximately 10% of cases.46 Common MRI perfusion techniques in clinical practice include DSC, DCE, and ASL, each with specific advantages and limitations. DSC, based on T2-weighted imaging following rapid contrast injection, effectively reflects tumor perfusion but is prone to artifacts from metal, blood, and air. DCE, based on T1-weighted contrast-enhanced imaging, is less artifact-prone but requires advanced post-processing software, limiting widespread use. ASL does not require gadolinium injection but has lower spatial resolution, which can limit clinical utility.
Additionally, susceptibility-weighted imaging (SWI) has been used to observe microbleeds and assess radiation-induced microvascular changes. Peters et al.47 found that children develop punctate SWI lesions more rapidly and earlier than adults. In a longitudinal study of pediatric patients treated for brain tumors with proton therapy, the cumulative incidence of radiation-induced cerebral microbleeds, as detected by SWI, increased progressively over time—reaching 43% at one year and 83% by five years post-therapy. The occurrence of cerebral microbleeds correlated significantly with higher radiation doses, greater irradiated brain volume, and younger age at treatment (p < 0.01).48 MRS is a non-anatomical imaging technique that quantifies metabolite concentrations within specific brain regions. MRS enables non-invasive detection of metabolic changes, allowing potential quantification of radiation-induced neurotoxicity. In the healthy brain, metabolites such as N-acetylaspartate and choline show distinct peaks.49 Among 242 patients who underwent both MRS and PET-computed tomography (CT) examinations, the diagnostic accuracy of MRS was 81.8%, significantly higher than that of PET-CT (42.9%).50 Changes in neuronal and glial cell populations alter intracranial metabolite concentrations.51 The N-acetylaspartate/creatine ratio is higher in regions of radiation necrosis than in tumor regions, while choline/N-acetylaspartate and choline/creatine ratios are higher in recurrent tumors. Reduced choline levels, along with potentially elevated lipid and lactate signals, suggest radiation necrosis.52 However, the long scan times required for accurate assessment limit the routine clinical use of MRS.
Deoxyhemoglobin is more paramagnetic than oxyhemoglobin, serving as a natural contrast agent. When vascular damage following radiotherapy causes imbalances between oxygen uptake and cerebral circulation, MRI sequences sensitive to magnetic field inhomogeneity can detect signal changes around cortical vessels. This is known as blood oxygenation level-dependent (BOLD) contrast. BOLD allows functional localization studies without contrast agents while providing high spatial resolution. Signal changes theoretically depend on intracranial blood oxygenation, blood flow, hematocrit, and tissue oxygen uptake, with blood flow being the primary determinant. BOLD functional MRI includes task-based and resting-state fMRI. Due to its ability to non-invasively measure hemodynamic changes, BOLD fMRI enables assessment of local neuronal and synaptic activity and, with high spatial and temporal resolution, has been widely used in neurological, psychiatric, and psychological research.53,54 This technique may also be applicable for detecting RIBI.
In cases of cerebral radiation necrosis, conventional MRI typically shows ring-enhancing lesions at the treatment site with surrounding edema, which are non-specific and may also occur in tumor progression.55 Diagnostic uncertainty for radiation necrosis based on conventional imaging alone can reach up to 15%.56 With the increasing use of immunotherapy, this uncertainty is further heightened, as pseudoprogression related to immune responses is often indistinguishable from true tumor progression using contrast-enhanced MRI alone.55 Therefore, additional imaging modalities are needed to supplement conventional methods for accurate evaluation. Based on the above content and previous studies, we summarized the general diagnostic and treatment workflow for MRI in RIBI (Fig. 2).
Radiomics and artificial intelligence analysis
In recent years, radiomics combined with machine learning has emerged as a forefront approach for diagnosing RIBI. By extracting high-dimensional quantitative imaging features (such as intensity, texture, and shape), classification models are established to assist in distinguishing tumor recurrence from radiation necrosis. In 2023, Salari et al.60 constructed a random forest model based on T1+C and FLAIR sequence images, achieving over 90% classification accuracy in the validation cohort, providing technical support for noninvasive and automated diagnosis of RIBI. Moreover, deep learning models have shown promising prospects in predicting and dynamically evaluating RIBI.
The integration of multimodal imaging technology has allowed the assessment of RIBI to evolve from traditional structural evaluation toward comprehensive analysis encompassing function, metabolism, microstructure, and inflammatory status. Each imaging modality possesses distinct advantages, and their rational combined application facilitates earlier detection, improved differential diagnosis, and precise guidance for clinical intervention strategies.