Washout maps could tell brain tumor regrowth from radiation damage after treatment

For patients with malignant brain tumors, the most nerve-wracking moment often comes months after treatment ends. A routine follow-up MRI shows a new, bright, contrast-enhancing lesion in the brain, and suddenly everyone faces the same question: is the tumor growing back, or is this just radiation damage? A team of German researchers now reports that a novel imaging technique, built on how quickly contrast agent leaves a lesion, could help answer that question with surprising clarity. In an exploratory study published in BMC Cancer, the so-called washout ratio separated recurrent tumor from radiation necrosis with an area under the curve of 0.86, outperforming the perfusion and diffusion measures that radiologists currently rely on.

The problem this technique addresses is one of the oldest in neuro-oncology. After surgery, radiotherapy and chemotherapy, malignant brain tumors such as glioblastoma are monitored with contrast-enhanced T1-weighted MRI, the cornerstone of the widely used Response Assessment in Neuro-Oncology, or RANO, criteria. According to these criteria, changes in contrast enhancement are key indicators of tumor progression. But radiation necrosis and pseudoprogression, therapy-related changes that involve no viable tumor, can produce strikingly similar patterns of enhancement. This overlap substantially limits the specificity of conventional MRI and complicates reliable response assessment, because misclassification can lead either to premature discontinuation of effective therapy or to delayed treatment of a true recurrence.

Current response frameworks emphasize caution. RANO 2.0 recommends that post-radiotherapy MRI serve as the baseline study and that findings suggestive of progression within three months of completing radiotherapy generally be confirmed on follow-up imaging before being labeled as true progression. Advanced techniques such as MR perfusion, MR spectroscopy, diffusion-weighted imaging and amino acid PET can help, but each has drawbacks. Perfusion results depend heavily on scanner settings, post-processing software and reader strategy, and PET requires specialized infrastructure and access to radiotracers that many centers lack. The new approach, developed at Otto-von-Guericke University Magdeburg, takes a different route: it interrogates the delayed contrast kinetics of a lesion rather than its vascularity or metabolism.

The biological rationale comes from experimental work showing that high-grade gliomas tend to reach peak enhancement around three to eight minutes after contrast injection. Building on this, the Magdeburg team optimized their imaging protocol by acquiring 3D T1-weighted sequences approximately five and twenty-five minutes after contrast administration. The two image sets are co-registered using the FMRIB FLIRT algorithm, intensity-normalized, and subtracted voxel-wise by a Python script that retrieves both sequences from the hospital’s picture archiving system. The resulting difference map is displayed in a red-green color scheme, with red assigned to washout effects and green to delayed or late enhancement. Areas of rapid contrast turnover are thought to correspond to viable, hypervascular tumor, whereas areas of contrast accumulation represent treatment-related tissue injury, scarring or necrosis.

To test whether this approach could distinguish true progression from radiation necrosis, the researchers retrospectively identified patients with histologically confirmed glioblastoma, WHO grade 3 or 4 astrocytoma, primary CNS lymphoma or brain metastasis who had undergone radiochemotherapy and participated in the clinic’s MRI protocol, including MR perfusion and rapid washout imaging, between January 2024 and December 2025. Only patients with new contrast-enhancing lesions on follow-up MRI and measurable enhancing volumes exceeding 125 cubic millimeters were included. The final cohort comprised 33 patients with a mean age of 58.8 years, including 25 with glioblastoma, four with astrocytomas, one with CNS lymphoma and three with brain metastases.

Diagnostic confirmation was rigorous. Histopathological assessment provided the reference standard in 20 patients, while longitudinal clinical and MRI follow-up of at least six months confirmed the diagnosis in 13 patients, with fluorodeoxyglucose-PET providing additional support in two cases. The final diagnoses were 23 cases of tumor recurrence, four cases of mixed pathology, defined as radiation necrosis with scattered active tumor cells, and six cases of pure radiation necrosis. Notably, no cases of pseudoprogression were identified; all therapy-associated changes in the cohort represented radiation necrosis, although pseudoprogression could not be excluded as a component of the mixed-pathology cases.

The central result is a striking difference in contrast kinetics between the diagnostic groups. The mean washout ratio was 31 plus or minus 18 percent for progressive disease, 35 plus or minus 20 percent for mixed pathology, and only 10 plus or minus 7 percent for radiation necrosis. Receiver operating characteristic analysis identified an exploratory threshold of less than 23 percent for the washout ratio, achieving a sensitivity of 100 percent and a specificity of 67 percent for detecting radiation necrosis, with an area under the curve of 0.86. An alternative threshold of 15 percent traded sensitivity for specificity, yielding 75 percent and 85 percent respectively. By comparison, MR perfusion performed considerably worse in this cohort: the sensitivity of perfusion for detecting radiation necrosis was 83 percent, but specificity was only 52 percent, and the area under the curve for the hyperperfused lesion volume was just 0.64. Diffusion-weighted imaging, measured as the volume of significant reductions in apparent diffusion coefficient values, also showed weaker discrimination, with an area under the curve of 0.61.

The researchers also combined all three markers in a multivariable logistic regression model integrating ADC, washout ratio and perfusion parameters. The combined model achieved an area under the curve of 0.89 for differentiating radiation necrosis from active tumor cells, with a model probability threshold of 0.16 providing 100 percent sensitivity and 72 percent specificity. This supports the study’s key conclusion that diagnostic discrimination may be strengthened by combining relative cerebral blood volume with washout ratio analysis rather than relying on a single imaging marker. Reliability measurements added confidence in the technique itself: washout volume segmentation showed excellent interrater agreement with an intraclass correlation coefficient of 0.92, and intra-rater reliability after a four-week interval was nearly perfect at 0.99, notably higher than the reliability of perfusion-derived and diffusion-derived volumes.

The technique belongs to a family of methods known as Treatment Response Assessment Maps, or TRAMs, and contrast clearance analyses, which are generated from two post-contrast T1-weighted acquisitions at different time points. Because these maps rely on delayed T1 contrast behavior rather than T2-star susceptibility signal changes, they are less affected by artifacts from blood products, postoperative material or calcification, which can degrade dynamic susceptibility contrast perfusion maps, particularly in postoperative glioblastoma cavities or hemorrhagic lesions. A histology-validated study of TRAMs previously demonstrated that regions of contrast clearance identified active tumors with 100 percent sensitivity and 92 percent positive predictive value. The maps are also visually intuitive, displaying contrast kinetics in a color scheme that radiologists can read at a glance, which may improve clinical usability when conventional MRI and perfusion findings are inconclusive.

The authors are careful to frame their findings as hypothesis-generating rather than practice-changing. The retrospective design, the small cohort of 33 patients, the heterogeneous histologies and the incomplete histopathological confirmation all increase the risk of overfitting, and the identified thresholds should not be regarded as validated clinical cutoffs. A single rCBV threshold of greater than 1.5 was applied across different scanners and field strengths, which may have affected perfusion classification. The researchers also note that washout mapping and FET-PET assess different biological processes, delayed contrast kinetics versus amino acid transport and tumor metabolism, and could therefore provide complementary information, though direct comparative studies are needed. Still, the conclusion is clear: washout and late enhancement provide additional diagnostic information that may improve lesion characterization in glioblastoma, and a combined approach integrating washout, late enhancement, ADC values and perfusion imaging appears to be a promising strategy for future diagnostic assessments, pending validation in larger and independent cohorts.

Subject of Research: Neuroscience

Article Title: Late enhancement and washout maps for distinguishing tumor progression from radiation necrosis in malignant brain tumors: an exploratory study

Article References: Khadhraoui, E., Schmidt, L., Klebingat, S., Sandalcioglu, I. E., Schwab, R., Arndt, P., Behme, D., Stein, K.-P., & Müller, S. J. (2026). Late enhancement and washout maps for distinguishing tumor progression from radiation necrosis in malignant brain tumors: an exploratory study. BMC Cancer, 26(1), Article 1238. https://doi.org/10.1186/s12885-026-17170-0

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17170-0

Keywords: MRI, washout maps, radiation necrosis, glioblastoma, tumor progression, contrast enhancement, MR perfusion, RANO criteria, neuro-oncology, brain tumors, TRAMs, diagnostic imaging