基于两级无监督学习的青少年腰椎骺环发育分期体系的建立OA

Objective:To overcome the limitations of traditional methods for assessing adolescent lumbar epiphyseal ring maturity,which rely on subjective empirical rules,this study aims to apply a data-driven approach to establish an objective new staging system for development and utilize this system to explore the developmental patterns of lumbar epiphyseal rings in the Chinese population.Methods:A retrospective analysis was conducted on lumbar CT images from 522 healthy individuals aged 6 to 30 years old.A two-level unsupervised learning framework was employed:K-Modes clustering was used to establish the epiphyseal ring maturation(ERM)staging at the epiphyseal ring level,and K-Means clustering was used to identify overall maturity patterns at the individual level.The validity was verified by comparing the method with conventional methods and correlating it with age,along with exploratory analyses of developmental windows,maturation paths,and cranio-caudal sequences.Results:An objective system comprising 5 ERM stages and 4 overall maturity patterns was successfully established.The new system showed a high correlation with age(ρ=0.890,P<0.001)and outper-formed traditional methods in identifying developmental subtypes.Two key developmental window periods were qua-ntified(initiation period Profile A,greatest individual differ-ences;fusion period Profile C,longest duration).The matura-tion path of"synchronous initiation→posterior leading→anterior closure"was revealed,and a developmental sequence from caudal to cranial was confirmed(P=0.045).Conclu-sion:The new staging system,based on a two-level unsupervised learning approach,can objectively and accurately assess adoles-cent lumbar maturity,revealing complex developmental patterns and providing more reliable imaging tools for clinical assessment and risk identification.