While cone beam CT (CBCT) is commonly used in musculoskeletal imaging of the extremities, its application in spinal imaging has been restricted by small field-of-view (FOV) coverage. Recent advancements in gantry-based CBCT systems promise to enable comprehensive imaging of the spinal column. This study aimed to evaluate the performance of a novel gantry-based, multi-scan CBCT system for spinal imaging with complete anatomic coverage and compare it to energy integrating (EI)CT and photon counting (PC)CT using dose-matched protocols. An anthropomorphic torso phantom was used to simulate human anatomy. Gantry-based CBCT scans of the thoracolumbar spine were performed using different presets (low-dose, enhanced, best quality), while EICT and PCCT scans followed dose-matched clinical protocols. Qualitative image analysis was assessed by three blinded readers using a 4-point Likert scale, and quantitative analysis was conducted using global noise level (GNL) measurements. CBCT achieved diagnostic-quality imaging for the thoracic and lumbar spine, particularly with "best" and "enhanced" presets. Subjective image quality was highest for PCCT, followed by EICT and CBCT. CBCT demonstrated lower GNL than EICT, nearing PCCT levels. However, high radiation doses (5 mGy) were required for CBCT imaging of the upper thoracic spine (Th1-Th6) due to anatomical complexity, while low doses (0,5 mGy) sufficed for the lower thoracolumbar spine (Th7-S1). Gantry-based CBCT was able to generate diagnostic-quality images of large spinal regions at relatively low radiation doses in a phantom setting, although the upper thoracic spine (above Th6) required higher doses. The overall subjective image quality remained below EICT and PCCT.

Background: The aim of this paper is to develop a clinically useful benchmark signature of deep learning (DL)-based MRI segmentations and objectively predict user satisfaction based on a weighted combination of multiple performance metrics. Methods: This post hoc analysis of a prospective study analyzed MRI data from 68 patients (29.4 ± 5.6 years; 45% female) acquired during a randomized clinical trial. Fat fraction maps of axial Dixon MRI were used to segment six classes of the thigh. Two radiologists qualitatively scored DL-based segmentations using a 5-point Likert scale. A hybrid deep learning model (MobileNetV2 and DINOv2) was developed to predict Likert scores directly from MRI slices and segmentation maps. Additionally, a weighted combination of quantitative metrics (Dice score, Hausdorff distance, Jaccard index) was determined to best correlate with the Likert scores. Statistical analyses included Cohen’s kappa, Spearman correlation, and regression model performance (MAE, RMSE, ROC_AUC). Results: Interreader agreement for Likert scores was near-perfect (κ = 0.82). The DINOv2 model achieved lower MAE (0.2343) and RMSE (0.1129) than MobileNetV2 (0.3488 and 0.431, respectively) in predicting Likert scores. The weighted metric combination model, using Lasso regression, predicted Likert = 5 with 84.1% accuracy, 100% recall, and ROC_AUC of 0.824. Individual metrics showed weak-to-moderate correlation with Likert scores, with Dice and Jaccard indices for extensor intramuscular fat exhibiting the highest correlation (r = 0.37 and r = 0.34, respectively). Conclusions: The proposed DL-based model accurately predicted radiologist-assigned Likert five scores, while the weighted metric combination model aligned closely with user satisfaction, offering a practical alternative to manual validation for clinical adoption of automated segmentation tools.

Background: Bone marrow adipose tissue (BMAT) is a metabolically active fat depot implicated in skeletal aging and systemic metabolism. While chemical shift-encoding MRI enables noninvasive quantification of bone marrow fat fraction (BMFF), current knowledge of vertebral BMAT is largely limited to small cohorts or restricted anatomical regions. Purpose: To characterize the spatial distribution of vertebral bone marrow fat fraction across the spine and determine region-specific associations with age and sex in a population-based cohort. Materials and methods: In this retrospective study, MRI data from 82,920 UK Biobank participants recruited between 2006 and 2010 were analyzed. The baseline case subset ( n = 81,318) had a mean age of 54.8 ± 7.6 years, and 42,113/81,318 (51.8%) were female. Automated vertebral body segmentation enabled voxel-wise BMFF quantification using Dixon MRI. Vertebra- and region-level BMFF metrics were derived. Associations with age and sex were assessed using region-specific linear regression. Results: Vertebral BMAT demonstrated a non-linear cranio-caudal pattern. Mean BMFF was ~ 39% in the cervical, ~ 55% in the thoracic, ~ 45% in the lumbar spine, and ~ 70% in the sacrum. BMAT increased with age in all regions (all p < 0.001; increase per decade: cervical 3.33%, thoracic 3.41%, lumbar 2.95%, sacral 3.83%). Females exhibited higher BMFF than males across all regions (all p < 0.001). The overall spatial pattern remained consistent across age groups and sexes. Conclusion: Vertebral BMAT exhibits a structured, non-linear spatial distribution with region-specific age- and sex-dependent patterns. This population-based atlas provides normative reference data supporting region-specific BMAT assessment in skeletal and metabolic disease.