Mustafa Mahmut BARIS, Ekrem Anil SARI, Tugce TOPRAK, Serap SARI, Yüksel OLGUN, Enis Alpin GUNERI, Nuri KARABAY, Mustafa Alper SELVER
Neurological Sciences and Neurophysiology - 2026;43(1):24-30
Background: Otosclerosis is an osteodystrophy of the otic capsule that can result in progressive hearing loss. Computed tomography (CT) is an effective method for assessing otosclerosis. Previous investigations have focused on density measurements on a single cochlear slice; however, otosclerosis can affect the entire otic capsule. Objectives: The purpose of this study is to propose a new diagnostic method for otosclerosis by performing measurements on the whole pericochlear area, providing a diagnostic framework that better reflects the pathophysiology of the disease. Methods: Between March 2017 and October 2023, surgically proven otosclerosis patients with previously performed temporal CT investigations were collected. Volumetric annotation and isolation of the pericochlear area were performed on CT images using Myrian software. The gray-level co-occurrence matrix of annotated regions was computed in two-dimensional (2D) and 3D. Principal component analysis (PCA) is applied. All calculations are carried out in Matlab 2021b(R). Results: The results show that histogram and texture-based features fail to differentiate the patient and control groups in both 2D and 3D. However, the application of the proposed patch-based extraction together with the use of PCA demonstrated that 99.99% of the data variability can be represented by the first two principal components. Consequently, the performance in terms of correct classification, selectivity, and specificity percentages is calculated as 98.61% +/- 1.32%, 97.34% +/- 3.47%, and 97.34% +/- 2.32%, respectively. Conclusions: In this study, our findings demonstrate that with the proposed volumetric analysis, robust discrimination between affected patients and the control group was achieved.
