基於OCTA與LASSO機器學習演算法之兒童阻塞性睡眠呼吸中止術後殘留風險列線圖預測模型構建與驗證

// Construction and validation of a nomogram prediction model for postoperative residual risk of pediatric obstructive sleep apnea based on OCTA and LASSO machine learning algorithm

本研究結合視網膜光學相干斷層血管造影（OCTA）、臨床、生化及多導睡眠檢查（PSG）數據，預測兒童腺樣體及扁桃腺切除後殘餘阻塞性睡眠呼吸中止症（OSAS）｜四大關鍵指標包括：腺樣體A/N比率、呼吸暫停低通氣指數、超敏C反應蛋白（hs-CRP）及深層毛細血管叢血管長度密度（DCP-VLD），其中DCP-VLD為獨立保護因子｜模型準確度高（AUC=0.845），顯示眼底微血管灌注減少是殘餘OSAS新穎獨立標誌，具臨床風險評估價值。

// This study aimed to predict residual obstructive sleep apnea syndrome (OSAS) in children after adenotonsillectomy using retinal OCTA microcirculatory parameters combined with clinical, biochemical, and PSG data. A nomogram was developed using four key predictors: adenoid A/N ratio, apnea-hypopnea index, hs-CRP, and deep capillary plexus vessel length density (DCP-VLD), with DCP-VLD identified as an independent protective factor. The model showed strong predictive accuracy (AUC=0.845) and good clinical utility. Reduced macular DCP perfusion is a novel, independent marker for residual OSAS, supporting its use in preoperative risk assessment.

Published 2026年7月25日