科研速览 · Science Skim继续刷下去 · Keep skimming →
◆ Spine deformity2026-09-08

Research priorities in Spine Deformity: a machine learning-based topic analysis of the journal's first decade (2013-2026).

Rohan A Phadke, Samer G Salman, Nathan J Lee

一句话结论 · In one sentence

Transformer-based topic modeling provides a data-driven map of Spine Deformity's first decade, revealing rapid growth in PSF outcomes research, frailty-aware perioperative care, registry-based outcomes analysis, and AI/LLM applications, alongside maturation of historical sagittal alignment and biomechanical research.

原始摘要(英文原文)· Original abstract
PURPOSE: Spine Deformity is the official journal of the Scoliosis Research Society (SRS) and the leading publication dedicated to scoliosis, deformity, and surgical correction. We applied BERTopic, a transformer-based machine learning topic-modeling framework, to characterize the evolution of research priorities in Spine Deformity from 2013 to 2026. METHODS: All English-language abstracts published in Spine Deformity from January 2013 through May 2026 (n = 1851 after quality filtering) were embedded using PubMedBERT, and then clustered with uniform manifold approximation and projection plus hierarchical density-based spatial clustering of applications with noise, and annotated via class-based term frequency-inverse document frequency. Temporal trends were tested with ordinary least-squares regression and Benjamini-Hochberg false discovery rate correction (α = 0.05). RESULTS: BERTopic identified 29 distinct research topics (noise = 18.4%). Six topics were emerging: postoperative posterior spinal fusion (PSF) outcomes and upper instrumented vertebra (UIV) selection; database studies of length-of-stay (LOS) and adverse events; frailty and perioperative risk; PSF in cerebral palsy and neuromuscular scoliosis; online patient education using large language models (LLMs); and socioeconomic disparities in deformity care. Two topics were cold: sagittal alignment and spinopelvic parameters in AIS, and biomechanical pedicle screw-rod testing. Although no standalone machine learning (ML) cluster emerged, ML and artificial intelligence (AI) methods appeared in 30 abstracts (1.6%), concentrated in radiographic measurement, risk prediction, and LLM-based patient education. CONCLUSION: Transformer-based topic modeling provides a data-driven map of Spine Deformity's first decade, revealing rapid growth in PSF outcomes research, frailty-aware perioperative care, registry-based outcomes analysis, and AI/LLM applications, alongside maturation of historical sagittal alignment and biomechanical research.
读原文 · Read the paper ↗

AI 追问PRO

登录后使用 AI 追问

讨论区

登录后参与讨论

相关论文 · Related

Research priorities in Spine Deformity: a machine learning-based topic analysis of the journal's first decade (2013-2026). — 科研速览 Science Skim