The research team set out to answer a deceptively simple question: when the same brain metastasis close to the skull is outlined on different imaging windows and modalities, how much do the resulting radiation dose distributions actually differ? To address it, the investigators performed a retrospective analysis of 245 lesions from 132 patients who had undergone radiotherapy. The lesions were divided among three treatment approaches: 88 treated with whole-brain radiation therapy, 67 with local radiation therapy, and 90 with a simultaneous integrated boost, a modern intensity-modulated technique that delivers a higher dose to visible tumors while treating the whole brain or surrounding tissue at a lower level simultaneously.

The methodological design is what gives the study its power. In the original clinical plans, gross tumor volumes had been delineated on contrast-enhanced brain-window CT images, a window setting optimized for brain parenchyma rather than bone. The researchers designated these volumes as GTV_Brain. They then rigidly registered contrast-enhanced MRI to the planning CT and created two additional delineations on the fused images: GTV_Skull, based on contrast-enhanced bone-window CT images that emphasize the bony structures of the skull, and GTV_MRI, based on contrast-enhanced T1-weighted MRI, which is widely regarded as the most sensitive modality for visualizing brain metastases because gadolinium contrast leaks across the disrupted blood-brain barrier of tumor tissue.

From each of the three gross tumor volumes, the team generated corresponding planning target volumes by applying a uniform three-millimeter expansion in all directions, producing PTV_Brain, PTV_Skull, and PTV_MRI. Critically, they also used Boolean operations to define MRI-only regions—the portions of the MRI-based target that extended beyond either the brain-window or bone-window CT-based targets. These regions represent tumor tissue that CT-based delineation would have missed entirely. Rather than recalculating new treatment plans, the researchers kept each patient’s original clinical plan and its dose distribution unchanged, extracting dose-volume parameters directly from the original radiotherapy dose files. This elegant approach isolates the pure effect of delineation differences on dose coverage, without confounding from plan optimization.

The primary dosimetric endpoint was D98%, the dose received by at least 98 percent of the target volume—a standard metric for confirming that nearly the entire target gets adequate coverage. The team normalized this value to the prescription dose, defining nD98% below 1.00 as a below-prescription-dose event and nD98% below 0.95 as a marked undercoverage event, a threshold beyond which a substantial portion of the tumor could harbor viable, untreated disease. Across all three treatment modalities, the pattern was strikingly consistent: PTV volumes always followed the order PTV_MRI greater than PTV_Skull greater than PTV_Brain, with the differences highly statistically significant at P less than 0.001.

The dose consequences were equally consistent and clinically sobering. D98% was highest for PTV_Brain, intermediate for PTV_Skull, and lowest for PTV_MRI. Compared with the brain-window and bone-window CT targets, the MRI-based target showed significantly lower D98% values, with reductions of 0.68 Gy and 0.58 Gy respectively in the whole-brain radiation therapy group, 1.96 Gy and 1.90 Gy in the local radiation therapy group, and 3.16 Gy and 2.82 Gy in the simultaneous integrated boost group. In other words, the more conformal and dose-escalated the technique, the greater the penalty for missing tumor tissue during delineation. A technique that concentrates radiation tightly around a drawn target is unforgiving of any tissue left outside that target.

The incidence figures translate these Gy-level differences into patient-level risk. Below-prescription-dose events for the MRI-based target occurred in 22.73 percent of lesions treated with whole-brain radiation therapy, 35.82 percent with local radiation therapy, and a striking 62.22 percent with simultaneous integrated boost. Marked undercoverage—nD98% falling below 0.95—affected 6.82 percent, 17.91 percent, and 31.11 percent of lesions in those same groups. Most alarming of all, within the MRI-only regions themselves, marked undercoverage reached 41.11 percent and 42.22 percent in the simultaneous integrated boost group, depending on which CT-based reference target was used. Nearly half of the tumor tissue visible only on MRI sat in low-dose zones that the original plans never intended to treat at full prescription.

Why does this matter biologically? A region of gross tumor that receives less than the prescription dose—and especially less than 95 percent of it—represents a geographic miss, a pocket of viable cancer cells that survives the treatment course. For brain metastases, such residual disease can lead to local recurrence, neurological deterioration, and the need for repeat radiation or surgery, all of which carry substantial morbidity. The skull-adjacent location compounds the problem: lesions hugging the inner table of the skull are precisely where contrast enhancement on CT is hardest to distinguish from adjacent bone and dural structures, and where the steep dose gradients of modern conformal techniques leave the least margin for error. The study’s data show that bone-window CT, while better than brain-window CT at capturing tumor extent near bone, still falls short of MRI.

The technical explanation lies in the physics of the two modalities. CT excels at depicting electron density, which is why it remains the backbone of dose calculation, but its soft-tissue contrast is limited, and beam-hardening artifacts near bone can obscure subtle enhancement. T1-weighted MRI after gadolinium injection, by contrast, highlights areas where the blood-brain barrier has broken down, revealing tumor margins that are invisible or ambiguous on CT. When the researchers fused the two datasets and compared delineations, the MRI-referenced targets were consistently larger, and the extra volume—the MRI-only regions—tended to lie in areas that the original plans had deliberately spared, such as regions near the scalp, skull, and normal brain that planners had constrained to keep doses low.

The authors’ conclusion is direct and actionable: for brain metastases close to the skull treated with local radiation therapy or simultaneous integrated boost, routine fusion of contrast-enhanced MRI with the planning CT is recommended for gross tumor volume delineation to reduce the potential risk of target geographic miss. The study, approved by the ethics committee of the Affiliated Cancer Hospital of Shandong First Medical University and conducted with written informed consent from all patients, was supported by research programs in Shandong Province and the Xinjiang Uygur Autonomous Region. As intensity-modulated radiotherapy, volumetric modulated arc therapy, and stereotactic techniques continue to push dose gradients ever steeper, this work serves as a quantitative reminder that the sharpest treatment plan is only as good as the contour it is built on—and that for tumors pressed against the skull, the contour drawn without MRI may simply be the wrong one.

Subject of Research: Dosimetric comparison of CT-based and MRI-based target delineation in radiotherapy for brain metastases close to the skull

Article Title: Dosimetric analysis of CT/MRI-based target delineation differences in radiotherapy for brain metastases close to the skull

Article References: Sun, S., Gong, G., Zhao, L., & Yin, Y. (2026). Dosimetric analysis of CT/MRI-based target delineation differences in radiotherapy for brain metastases close to the skull. BMC Cancer. https://doi.org/10.1186/s12885-026-17067-y

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17067-y

Keywords: brain metastases, radiotherapy, target delineation, MRI, computed tomography, dosimetry, D98%, gross tumor volume, planning target volume, simultaneous integrated boost, geographic miss, radiation oncology

The research, published in Veterinary Oncology, focused on canine acanthomatous ameloblastoma, or CAA, the most common odontogenic tumor in dogs, accounting for roughly ten percent of oral neoplasms in a species where oral tumors overall represent six to seven percent of all cancers. The gold standard treatment is surgical excision with margins of five to ten millimeters, which makes accurate tumor delineation on preoperative imaging critically important. If the tumor boundary drawn on a CT scan is wrong, the surgeon’s plan may be wrong too, and incomplete excision carries well-documented consequences: higher recurrence rates, reduced survival, and poorer overall prognosis.

To test the reliability of manual tumor outlining, the team selected ten confirmed CAA cases from the Cornell University Hospital for Animals CT database covering 2014 to 2021. Every case had been diagnosed by histopathology, and each scan included both non-contrast and contrast-enhanced images with bone and soft tissue reconstructions. The cohort comprised four Labrador retrievers and six other dogs of assorted breeds, with a mean age of 7.4 years. Nine tumors sat in the mandible, six rostrally and three caudally, while one arose from the left maxillary canine. Scans were acquired on a Toshiba Aquilion 16-slice scanner at 120 kilovolt peak, with pixel dimensions averaging 0.40 millimeters and slice thicknesses ranging from 0.5 to 2 millimeters.

Three board-certified specialists, two veterinary dentists and oral surgeons and one veterinary medical oncologist, independently contoured each of the ten tumors twice, in two randomized sittings separated by at least one week to reduce recall bias. Working in the open-source platform 3DSlicer, they could toggle between contrast and non-contrast datasets and use paintbrush or drawing tools on each transaxial slice. They received only crude tumor locations and a dental map, with no radiology reports or medical records, deliberately mimicking the conditions under which such contouring is often performed in practice without radiologist input. Each case took between 30 and 60 minutes, and no participant faced time restrictions.

The investigators then compared all the resulting contours using three complementary metrics. The Dice Similarity Index quantifies spatial overlap between two segmentations, ranging from zero for no overlap to one for perfect overlap. The Hausdorff Distance captures the worst-case error, measuring the maximum distance between the boundaries of two contours. Centroid positions, the arithmetic mean coordinates of all voxels in a segmented volume, reveal whether observers placed the tumor’s geometric center in the same spot. Volumes were also compared, with the median tumor volume across all trials measuring 2.46 cubic centimeters and a range from 1.01 to 9.46 cubic centimeters.

The headline finding was a striking split between global and local agreement. Correlation coefficients for centroid coordinates and tumor volumes showed excellent agreement both within and between observers, suggesting that everyone was capturing roughly the same mass in roughly the same place. Yet the boundary-level metrics told a more nuanced story. Intraobserver Dice values averaged 0.78 with a standard deviation of 0.11, indicating only moderate agreement even with oneself across repeated trials. Interobserver Dice values dropped to a mean of 0.69 with a standard deviation of 0.14, which the study’s classification scheme labels as poor agreement between different participants.

Hausdorff Distances reinforced that pattern. Within a single observer’s two trials, the mean worst-case boundary discrepancy was 0.80 millimeters, but between different observers it climbed to 1.21 millimeters, with a standard deviation of 0.66. Effect size analysis using Cohen’s d showed small, medium, and large effects for the three pairwise observer comparisons, and interobserver volume estimates were statistically different at P less than 0.001, with effect sizes ranging from 0.31 to 0.72. Analysis of variance also flagged significant differences in the centroid Z coordinate between one pair of observers and in volumes between another. Notably, the authors estimated that observer differences accounted for 15.6 percent of volume variability and 17.7 percent of Hausdorff Distance variability.

Why does this matter beyond veterinary clinics? Because manual segmentations are the training data for the radiomics and AI models now being built to diagnose, classify, and prognose tumors automatically. In human medicine, variations in gross tumor volume delineation can reach as much as twenty percent, and radiomics-based models have shown good diagnostic performance in head and neck cancers, canine small intestinal tumors, and renal tumors. If the ground truth used to train these algorithms is itself inconsistent, the models inherit that inconsistency. The Cornell team frames their work as a technical validation step, one of the internal, external, and technical validation processes required before AI models can be trusted in the clinic, and their reliability data can now serve as a benchmark for comparing segmentation methods and developing autosegmentation tools.

The authors are candid about their study’s limitations. Ten cases and three observers is a small sample, CT acquisition parameters varied across cases, slice thickness ranged up to 2 millimeters along the Z axis, and the cohort overrepresented rostral mandibular tumors and Labrador retrievers, factors that could matter when building radiomics models even if they likely had little effect on observer agreement itself. Time per case was self-reported rather than measured, so learning effects cannot be excluded, and the design deliberately departed from standard clinical workflow by withholding radiology and pathology reports. The team also notes that whether these minor contour inconsistencies would actually change surgical margins or outcomes cannot be determined without histopathological assessment of complete excisions, a question they hope larger studies will address.

Still, the takeaway is clear and broadly relevant: even experienced specialists tracing the same tumor on the same scans produce boundaries that overlap only moderately, even though the tumor’s center and size come out remarkably consistently. As radiomics and AI move from promising research tools toward clinical integration in both human and veterinary oncology, the study argues that standardized contouring protocols are no longer optional. Reliable segmentation, the authors conclude, is foundational, and quantifying exactly how much human hands disagree is the first step toward building machines that do it better.

Subject of Research: Observer variability in manual CT tumor delineation for canine acanthomatous ameloblastoma and its implications for radiomics validation

Article Title: Assessment of inter and intraobserver variability in manually delineated oral canine acanthomatous ameloblastoma from computed tomography scans

Article References: Drozd, M. E., Fiani, N., Sylvester, S., Khalil, M. K., Peralta, S., & Basran, P. S. (2025). Assessment of inter and intraobserver variability in manually delineated oral canine acanthomatous ameloblastoma from computed tomography scans. Veterinary Oncology, 2(1), Article 37. https://doi.org/10.1186/s44356-025-00052-1

Image Credits: AI Generated

DOI: 10.1186/s44356-025-00052-1

Keywords: radiomics, canine acanthomatous ameloblastoma, computed tomography, tumor segmentation, observer variability, Dice Similarity Index, Hausdorff distance, veterinary oncology, 3DSlicer, artificial intelligence, surgical margins, gross tumor volume