Beyond Radiologic Perception: A Pilot Quantitative and Radiomic Comparison between Zero-Echo-Time MRI and CT in Oncologic Metastatic Screening

Abstract

Introduction: Zero-Echo-Time (ZTE) and Ultrashort-Echo-Time (UTE) are MRI pulse sequences designed to image tissues with very short transverse relaxation times on short-T2 materials, which lose signal rapidly after excitation and are typically not visible on standard MRI scans. These sequences can increase bone contrast, producing images that resemble computed tomography (CT) by using ray-sum rendering, which sums signal intensities along a line to mimic X-ray imaging, and inverse-logarithm rescaling, which adjusts signal values using the inverse of the logarithm to enhance contrast. Previous ZTE publications often employ perception analysis, evaluating images based on expert observer ratings, to validate the method as a CT-surrogate. We aim to quantitatively compare oZTEo, a GE Healthcare (GEHC)-developed implementation of ZTE MRI, with CT using radiomic analysis on normal bone.
Methods: Beginning in July 2025, ZTE/oZTEo was incorporated into routine musculoskeletal pelvic MRI protocols at our institution using 9 GEHC MRI scanners (2 at 3 Tesla, 7 at 1.5 Tesla). This prospective study includes 26 patients who underwent CT and MRI screening for bone metastases, with a time interval of less than 6 months between the modalities. MR and CT images were retrieved from the institutional database. Manual 3D volume-of-interest (VOI) segmentations were performed on healthy right hips, or on the left if the right was affected by pathology or hardware. Custom Python-based software extracted 107 quantitative radiomic features per VOI. Features were standardized using Z-score normalization. In cases where ZTE was acquired, gray-scale inversion was applied, with bright regions becoming dark and vice versa, to match CT style. In contrast, for commercial oZTEo acquisition, gray-scale inversion was already automatically applied during product reconstruction; therefore, further gray-scale inversion was not needed. Statistical comparisons were performed for all 107 features between oZTEo (n = 26) and CT (n = 26). Manual measurements of multifocal cortical and marrow intensity were also obtained from all patients. Resulting intensity histograms from both modalities were standardized, aligned, and averaged across patients.
Results: Of the 26 patients, 11 (42%) were male and 15 (58%) were female, with a mean age of 50 years (range: 24–76). Twenty-two right and four left hip VOIs were segmented on both oZTEo and CT scans. The mean interval between MRI and CT was 2 months (range: 1 day to 6 months). Following gray-scale inversion of ZTE radiomics, all features were statistically similar to CT (p > 0.00047). Without this correction, ZTE features appeared as inverted images of the CT feature distribution. Mean cortex and marrow intensities from oZTEo were 1,653 ± 729 a.u. and 1,257 ± 691 a.u., compared to 1,148 ± 288 H.U. and 115 ± 110 H.U. for CT. Before standardization, the percentage differences between oZTEo and CT for the average histogram mean, skewness, kurtosis, 10th percentile, and 90th percentile were 324%, -54%, -100%, 1,528%, and 231%, respectively. Histogram area overlap was limited to 27%, with 42% and 3% overlap for cortex and marrow ranges. After standardization and histogram alignment, differences were markedly reduced: mean (0%), skewness (-10%), kurtosis (-20%), 10th percentile (3%), and 90th percentile (6%). Histogram area overlap increased to 95%, with cortex and marrow overlaps rising to 84% and 72%.
Conclusion: This pilot study provides initial quantitative and radiomic evidence that oZTEo and CT exhibit substantial objective similarity beyond perceptual visual assessment, supporting oZTEo as a clinically viable MRI alternative to conventional CT for bone evaluation. These promising pilot findings will be subsequently validated in a larger cohort, including cases with sclerotic and lytic bone lesions.