A Deep-Learning Model for Diagnosing Fresh Vertebral Fractures on Magnetic Resonance Images.

BACKGROUND: The accurate diagnosis of fresh vertebral fractures (VFs) was critical to optimizing treatment outcomes. Existing studies, however, demonstrated insufficient accuracy, sensitivity, and specificity in detecting fresh fractures using magnetic resonance imaging (MRI), and fall short in localizing the fracture sites. METHODS: This prospective study comprised 716 patients with fresh VFs. We obtained 849 Short TI Inversion Recovery (STIR) image slices for training and validation of the AI model. The AI models employed were yolov7 and resnet50, to detect fresh VFs. RESULTS: The AI model demonstrated a diagnostic accuracy of 97.6% for fresh VFs, with a sensitivity of 98% and a specificity of 97%. The performance of the model displayed a high degree of consistency when compared to the evaluations by spine surgeons. In the external testing dataset, the model exhibited a classification accuracy of 92.4%, a sensitivity of 93%, and a specificity of 92%. CONCLUSIONS: Our findings highlighted the potential of AI in diagnosing fresh VFs, offering an accurate and efficient way to aid physicians with diagnosis and treatment decisions.

Radiographic characteristics-based classification system for percutaneous endoscopic lumbar discectomy surgical approach selection in patients with L5-S1 disc herniation: a hierarchical clustering analysis.

Background: No classification system exists for aiding the selection of surgical approaches in L5-S1 disc herniation when undergoing percutaneous endoscopic lumbar discectomy (PELD). We aimed to identify radiographic subtypes to aid the selection of percutaneous endoscopic transforaminal discectomy (PETD) and percutaneous endoscopic interlaminar discectomy (PEID) in patients with L5-S1 disc herniation via unsupervised hierarchical clustering analysis. Methods: In this retrospective case-control study, we gathered 296 anteroposterior and lateral lumbar spine radiographs (dataset 1) from Tianjin Hospital between January 2016 and October 2021 for clustering analyses. Additionally, we analyzed 111 patients who underwent PEID or PETD for L5-S1 disc herniation at Tianjin Hospital from January 2016 to August 2022. We included patients with radicular leg pain or back pain associated with intra-canal disc herniation who failed in conservative treatments over 6 weeks. First, pair-wise Spearman correlation coefficients were calculated among plain radiographic metrics in dataset 1 to reveal the association among these radiographic metrics. Second, hierarchical clustering analysis was conducted to unsupervised cluster the plain films into several subtypes. Last, for each radiographic subtype, the intraoperative blood loss (IBL), operation time (OT), total operating room time (TORT) along with visual analogue scale (VAS) and Oswestry Disability Index (ODI) were compared between patients underwent PETD or PEID with age as covariates. Results: This study yielded 3 main findings: (I) iliac height (IH) was negatively correlated with intervertebral foramen width (IFW), intervertebral foramen height (IFH), and intertransverse height (ITH) (R=-0.50, -0.42, and -0.46, all P<0.001), ITH was positively correlated with IFW and IFH (R=0.40 and 0.53, all P<0.001); (II) 2 lumbosacral radiographic subtypes were identified via hierarchical clustering analysis; (III) relative to subtype 1, the patients identified as subtype 2 exhibited lesser IBL, shorter OT, and shorter TORT following PETD (t=2.92, P=0.006; t=2.65, P=0.012; t=3.17, P=0.003). Conclusions: The morphology pattern of the lumbosacral region affect the ease of different PELD procedures when performing percutaneous discectomy at the segment of L5-S1. Without considering the type of disc herniation, this classification system might aid spine surgeons in the selection of an appropriate surgical approach.