Background/Objectives: This study aimed to create a pipeline for standardized data extraction from lumbar-spine MRI radiology reports using a large language model (LLM) and assess the agreement of the extracted data with research-grade semi-quantitative scoring. Methods: We included a subset of data from a multi-site NIH-funded cohort study of chronic low back pain (cLBP) participants. After initial prompt development, a secure application programming interface (API) deployment of OpenAIs GPT-4 was used to extract different classes of pathology from the clinical radiology report. Unsupervised UMAP and agglomerative clustering of the pathology terms’ embeddings provided insight into model comprehension for optimized prompt design. Model extraction was benchmarked against human extraction (gold standard) with F1 scores and false-positive and false-negative rates (FPR/FNR). Then, an expert MSK radiologist provided comprehensive research-grade scores of the images, and agreement with report-extracted data was calculated using Cohen’s kappa. Results: Data from 230 patients with cLBP were included (mean age 53.2 years, 54% women). The overall model performance for extracting data from clinical reports was excellent, with a mean F1 score of 0.96 across pathologies. The mean FPR was marginally higher than the FNR (5.1% vs. 3.0%). Agreement with comprehensive scoring was moderate (kappa 0.424), and the underreporting of lateral recess stenosis (FNR 63.6%) and overreporting of disc pathology (FPR 42.7%) were noted. Conclusions: LLMs can accurately extract highly detailed information on lumbar spine imaging pathologies from radiology reports. Moderate agreement between the LLM and comprehensive scores underscores the need for less subjective, machine-based data extraction from imaging.

Effective management of non-specific chronic lower back pain (ns-cLBP) requires nuanced prescription decisions within evolving guidelines for conservative treatment. This study developed comprehensive LBP patient profiles from electronic medical records (EMR), integrating clinical charts (demographics, social determinants, diagnoses, medications) and radiology reports (MRI-confirmed diagnoses) to predict pharmacological management strategies. One-vs-one and one-vs-rest classification frameworks systematically evaluated treatment decisions across three prescriptions: no medication, NSAIDs, and opioids. Real-world complexity and heterogeneity in ns-cLBP management was reflected in modest yet clinically meaningful performance metrics (balanced accuracy = 0.58, AUC = 0.62, F1-score = 0.42). Chart-documented diagnoses marginally outperformed MRI-reported pathology as predictors, though this difference was within the range of variability, which suggests the importance of diagnoses informed by patient-reported symptoms in shaping treatment pathways. SHAP feature importance analysis identified consistent predictors (year_at_first_imaging) and variable factors (spinal_stenosis, disc_pathology, race_ethnicity, negative_psych_state, osteoarthritis_osteoarthrosis) in prescriptions, with higher associations observed in those with anxiety or depression, partnered individuals and females. By leveraging explainable AI, this study quantifies the interplay between biological and psychosocial drivers of prescribing decisions, offering a transparent, data-driven monitoring tool for understanding in chronic pain care. These findings demonstrate the potential of multi-modal EMR data and interpretable models to guide more personalized, equitable ns-cLBP management and opioid prescriptions.

A 2D U-Net was trained to generate synthetic T_1p maps from T₂ maps for knee MRI to explore the feasibility of domain adaptation for enriching existing datasets and enabling rapid, reliable image reconstruction. The network was developed using 509 healthy contralateral and injured ipsilateral knee images from patients with ACL injuries and reconstruction surgeries acquired across three institutions. Network generalizability was evaluated on 343 knees acquired in a clinical setting and 46 knees from simultaneous bilateral acquisition in a research setting. The deep neural network synthesized high-fidelity reconstructions of T_1p maps, preserving textures and local T_1p elevation patterns in cartilage with a normalized mean square error of 2.4% and Pearson’s correlation coefficient of 0.93. Analysis of reconstructed T_1p maps within cartilage compartments revealed minimal bias (−0.10 ms), tight limits of agreement, and quantification error (5.7%) below the threshold for clinically significant change (6.42%) associated with osteoarthritis. In an out-of-distribution external test set, synthetic maps preserved T_1p textures, but exhibited increased bias and wider limits of agreement. This study demonstrates the capability of image synthesis to reduce acquisition time, derive meaningful information from existing datasets, and suggest a pathway for standardizing T_1p as a quantitative biomarker for osteoarthritis.

Objective: Nonspecific low back pain (ns-LBP) is a widespread health challenge, and opioids were frequently prescribed upon first LBP presentation in the United States during the opioid epidemic. We assessed temporal changes in quarterly (Q) opioid prescriptions for new-onset ns-LBP from 2012 through 2024 at a California medical center amid evolving voluntary federal recommendations and mandatory state regulations. Methods: We analyzed opioid prescription rates (prescription within 6 months of diagnosis) for ns-LBP patients diagnosed in all departments and stratified by emergency department (ED), and we compared high- versus low-dose prescription (≥50 morphine mg equivalents/day). Change point detection identified transitions without policy priors, and interrupted time series regression quantified level changes and trend modifications. Results: Of 31 940 ns-LBP patients, 2077 (6.5%) received opioids (mean [SD] age, 51.0 [17.6] y; 45.3% male; 53.7% White). Change points occurred at 2015 Q2 and 2018 Q4. Prescribing odds declined 2.7% per quarter from 2012 Q1 through 2024 Q4 (95% CI, 0.4%-5.0%; P = .02). ED patients (16.5% prescribed opioids vs 5.8% non-ED) showed a 54.2% immediate reduction at 2015 Q2 (95% CI, 18.8%-74.2%; P = .008), declining from 27.4% to 12.2% before plateauing; non-ED patients showed a 42.8% reduction at 2018 Q4 (95% CI, 27.5%-54.9%; P < .001). High-dose prescribing remained unchanged among opioid recipients. Conclusion: Change points coincided with anticipatory adoption of the 2016 Centers for Disease Control and Prevention recommendations and 2018 California mandatory drug monitoring implementation. ED settings showed disproportionate reductions at 2015 Q2, consistent with knowledge diffusion of voluntary guidelines in settings with frequent opioid prescribing, while non-ED reductions coincided with immediate regulatory compliance at 2018 Q4. This distinction can guide future policy selection to address environment-specific adoption barriers.