Tamotsu Kamishima

Automated lumbar computed tomography (CT) Hounsfield unit (HU) measurement can support opportunistic osteoporosis screening, but heterogeneous Digital Imaging and Communications in Medicine (DICOM) inputs often disrupt automated workflows before measurement. We evaluated a local Qwen2.5-VL-7B vision-language model (VLM)-assisted front end for suitability routing, input planning, and HU-traceable DICOM-derived field-of-view/orientation adaptation upstream of an unchanged single-slice lumbar CT HU workflow. The VLM was restricted to routing and planning. In a 20-case FUJIFILM pilot cohort, observed HU output availability was higher with the front-end-assisted route than with direct processing (12/20 versus 5/20), while paired automatic-success cases showed excellent agreement with post hoc manual ImageJ (version 1.54g) measurements (ICC(A,1) = 0.997; MAE = 4.64 HU). Public DICOM stress testing further demonstrated improved workflow robustness across heterogeneous datasets. These findings support the use of a deployment-oriented front-end strategy to improve the auditability and robustness of automated lumbar CT HU measurement while preserving an unchanged downstream measurement workflow.

Rheumatoid arthritis (RA) frequently involves the wrist joints, where early inflammation may induce subtle changes in carpal spatial architecture. This study investigated whether baseline (BL) ultrasound inflammation is associated with longitudinal changes in AI-derived compactness metrics and whether these metrics provide information complementary to conventional wrist radiographic scores. This study included 159 patients with early RA who underwent BL and follow-up (FU) bilateral wrist radiography and BL wrist ultrasound. Four AI-derived centroid-based compactness metrics—root-mean-square-radius (RMSR), mean pairwise distance (MPD), median radius (R50), and 90th-percentile-radius (R90)—were calculated, and their annualized changes were evaluated in relation to conventional wrist SvdH scores and BL ultrasound inflammation, including gray-scale (GS), power Doppler (PD), GS+PD, and vascularity percentage (VS%). Associations were assessed using correlation and subgroup analyses. AI-derived compactness metrics at BL and FU, as well as their annualized changes, were not significantly associated with conventional wrist radiographic scores. In contrast, BL ultrasound inflammation was significantly associated with subsequent increases in compactness metrics. BL GS, PD, and GS+PD scores showed significant positive correlations with annualized changes in RMSR, MPD, R50, and R90 (r = 0.18–0.25, p < 0.05). Subgroup analyses similarly demonstrated greater increases in compactness metrics in patients with positive GS or PD activity. Annualized changes in AI-derived compactness metrics were associated with baseline ultrasound inflammation but not with changes in conventional wrist radiographic scores. These findings suggest that AI-derived compactness metrics may provide complementary quantitative information on global carpal spatial architecture in early RA.

Purpose Joint space narrowing (JSN) in rheumatoid arthritis (RA) can progress even during clinical remission. Conventional imaging lacks sensitivity for detecting early risks. This study evaluates AI-based analysis of bilateral joint-space asymmetry on hand radiographs to detect future JSN progression in radiographically normal RA. Materials and methods Forty-six radiographically normal RA patients underwent bilateral hand radiography and high-resolution peripheral quantitative computed tomography (HR-pQCT) at baseline, 12, and 24 months. Five joint space parameters were extracted from HR-pQCT. Gaussian mixture modeling (GMM) stratified patients into progressive and non-progressive phenotypes using 24 month data. An AI pipeline was developed to measure bilateral joint space asymmetry from baseline X-rays using deep learning–based segmentation and registration. Predictive models were constructed using logistic regression and evaluated via ROC curves. Results GMM identified 17 progressive and 29 non-progressive patients. HR-pQCT-derived intra-joint space width inconsistency (JSW.Inc) increased significantly over time in progressors (p = 0.0337). At baseline, HR-pQCT-based JSW.Inc showed fair predictive power (AUC = 0.696). AI-derived asymmetry at PIP3 and MCP3 joints showed stronger baseline prediction (AUCs = 0.739 and 0.714, respectively). By month 12, predictive performance improved across models (AUCs up to 0.836 for HR-pQCT and 0.748 for AI-based metrics). Conclusions AI-assisted analysis of bilateral joint space asymmetry from hand radiography may provide an accessible method for detecting radiographic JSN progression, with prognostic findings that should be interpreted as exploratory and hypothesis-generating. This method complements conventional imaging tools and may support future investigations into early risk stratification.

Routine computed tomography (CT) provides an opportunity for opportunistic vertebra-aware analysis beyond its original acquisition purpose. In this work, we study the engineering feasibility of transforming routine post-fracture lumbar CT into a compact structured case summary, rather than producing only a single black-box prediction. We propose a vertebra-aware 3D multi-task learning framework that jointly performs vertebral segmentation, density-related descriptor estimation, CT- and geometry-derived structure-aware descriptor estimation, vertebra-level fracture-related auxiliary modeling, derived case-level summary generation, and quality-control/uncertainty-aware output organization. The structure-aware descriptor is introduced as a framework-defined quantitative field for organizing density-related signal distribution and vertebral geometry on the current scan, not as a validated biomechanical measurement or intrinsic-strength estimator. Experiments on xVertSeg using five-fold case-level cross-validation show that the framework can generate coherent vertebra-wise structured outputs and support preliminary derived case-level discriminative analysis under limited supervision. To partially address the small-sample limitation, supplementary experiments on VerSe 2020 are conducted for external anatomical generalization and anatomical pretraining. The results indicate that VerSe-based pretraining improves segmentation stability and downstream descriptor consistency after xVertSeg fine-tuning. Overall, this study should be interpreted as an engineering proof-of-concept for report-oriented structured analysis of post-fracture lumbar CT, rather than as prospective prediction, biomechanical validation, or a clinically deployed decision-support system.

Manual single-slice lumbar quantitative computed tomography (QCT) depends on operator-driven slice selection and trabecular region-of-interest (ROI) placement. We developed a fully automated single-slice workflow for vertebral trabecular Hounsfield unit (HU) measurement that combines unsuitable-slice prescreening, dual-purpose segmentation, intra-patient slice-quality ranking, and a deterministic inner ROI rule. The pipeline includes an Eligibility Gate, QC-Envelope segmentation for broad, vertebral- and usability-preserving delineation, PairRank-Swin for best-slice selection, and dedicated trabecular segmentation for final quantitative analysis. In the independent external cohort, 4 cases were considered non-evaluable by both manual review and the pipeline, and 2 additional borderline-quality cases were manually measured but rejected by the pipeline; therefore, paired HU agreement analysis included 44 evaluable cases. Agreement remained high, with Pearson’s r = 0.987, Lin’s CCC = 0.985, mean bias −0.44 HU, and limits of agreement from −14.88 to +13.99 HU. Coverage was 84.1% within ±10 HU and 97.7% within ±15 HU. Ablation analysis showed that slice ranking and ROI erosion were the most critical components. In an open module-level baseline comparison, QC-Envelope segmentation substantially outperformed TotalSegmentator. This workflow provides high agreement with expert HU measurement while preserving reviewable intermediate outputs.