Glaucoma, one of the leading causes of blindness, often develops asymptomatically, necessitating early diagnosis and prediction of the progression rate of glaucomatous optic neuropathy (GON).
Purpose: To develop a classification model using machine learning methods for predicting the rate of GON progression, and to identify the most significant predictors of progression in patients with newly diagnosed early primary open-angle glaucoma (POAG).
Material and methods: The study included 59 patients (59 eyes) with early POAG, categorized into three groups based on the expert assessment of GON progression rate over a 36-month follow-up using dynamic morphofunctional evaluation. A classification model incorporating 35 clinical parameters, including optical coherence tomography (OCT) and OCT-angiography (OCT-A) data, was developed using partial least squares discriminant analysis (PLS-DA).
Results: Over the 36-month follow-up, slow GON progression was recorded in 21 patients, moderate in 18, and rapid in 20. The mean progression rates were -0.77±1.27%/year for visual field area, -1.21±1.48 µm/year for retinal nerve fiber layer (RNFL) thickness, and -1.23±1.77 µm/year for ganglion cell complex (GCC) thickness. The model demonstrated sensitivity of 90%, specificity of 95%, and efficiency of 92%. The most significant predictors of GON progression were mean vessel density in the deep vascular plexus of the macular region (wiVD_Deep), choriocapillaris dropout in the inferior-nasal peripapillary region, choroidal thickness in the fovea, and lamina cribrosa thickness.
Conclusion: The developed model effectively classifies patients based on the predicted progression rate of GON, which is important for individualized approach to glaucoma treatment planning.
扫码关注我们
求助内容:
应助结果提醒方式:
