基于高分辨率颞骨CT儿童感音神经性耳聋影像组学特征研究OA

Objective Based on high-resolution temporal bone CT images,this study aimed to develop and evaluate a radiomic model capable of distinguishing among three inner ear conditions,i.e.large vestibular aqueduct syndrome(LVAS),nonspecific profound sensorineural hearing loss(SNHL),and normal hearing,as well as the clinical significance of their morphological and textural differences.Methods A total of 278 ears examined between July 2015 and January 2025 were included,involving LVAS(n=72),nonspecific profound SNHL(n=66)and normal-hearing ears(n=140).Regions of the inner ear(semicircular canals,vestibule,cochlea,etc.)were manually segmented using the 3D Slicer,and radiomic features were extracted.LASSO regression was then applied for feature dimensionality reduction and selection.Multiple machine learning classifiers,including logistic regression,support vector machines,and random forests,were employed to perform multiclass classification among the three groups.Five-fold cross-validation was used to assess accuracy and area under the ROC curve(AUC).Morphological and textural differences across different anatomical regions were compared.Results Shape and texture features extracted from key anatomical structures(i.e.,the horizontal semicircular canal,vestibule,and cochlea)exhibited strong discriminatory power in differentiating among the three groups.Logistic regression and random forest models achieved favorable accuracy and AUC in the three-way classification task.Morphological analysis showed significantly increased vestibular and horizontal semicircular canal volume and surface area in LVAS,whereas bony structures in nonspecific profound SNHL were generally similar to normal-hearing ears,but with clear textural differences suggesting fundamentally different inner ear microstructural heterogeneity between the two conditions.Conclusions High-resolution temporal bone CT-based radiomics can effectively quantify and reveal subtle structural differences associated with distinct inner ear pathologies,providing objective evidence for distinguishing LVAS from other severe hearing losses and normal hearing.This approach holds promise for aiding the diagnosis,classification,and individualized treatment of congenital hearing loss,offering new perspectives and basis for precise clinical decision-making.