結合血液學和篩竇與上頷竇分析以進行鼻竇炎評估之演算法

Abstract

嗜酸性慢性鼻竇炎（ECRS）是一種嚴重影響患者生活品質的上呼吸道發炎性疾病。傳統診斷方法往往依賴侵入性組織活檢，限制了大規模篩檢的可行性。本研究提出了一種創新的鼻竇分割演算法BREATHE（Boosted Rhinosinusitis Evaluation Algorithm Through Hematology and Ethmoid-maxillary Analysis），目的在於克服現有ECRS診斷方法的限制。BREATHE演算法透過一系列步驟自動分析CT影像，包括骨骼提取、眼窩偵測和鼻竇分割。我們開發了有效的CT影像篩選方法，自動選擇最適合解釋的影像。此外，本研究也整合了血液學指標，特別是嗜酸性球與嗜中性球比值（EN ratio），以提高診斷準確性。實驗結果表明，BREATHE方法在ECRS診斷中表現出色。 E-M評分與EN ratio的結合效果最佳，顯著提高了診斷準確性。透過使用邏輯迴歸和隨機森林模型的整合方法，我們進一步提升了預測性能。本研究的主要創新點包括：提出不依賴大規模資料集的自動化鼻竇分割演算法、開發有效的CT影像篩選技術。這些創新為ECRS的早期診斷和精確治療提供了新的工具和見解，有望顯著改善患者的生活品質。

Eosinophilic Chronic Rhinosinusitis (ECRS) is an inflammatory disease of the upper respiratory tract that significantly impacts patients' quality of life. Traditional diagnostic methods often rely on invasive tissue biopsies, limiting the feasibility of large-scale screening. This study proposes an innovative sinus segmentation algorithm, BREATHE (Boosted Rhinosinusitis Evaluation Algorithm Through Hematology and Ethmoid-maxillary Analysis), aimed at overcoming the limitations of existing ECRS diagnostic methods. The BREATHE algorithm automatically analyzes CT images through a series of steps, including bone extraction, eye socket detection, and sinus segmentation. We developed an effective CT image screening method that automatically selects the most suitable image for interpretation. Additionally, this study integrates hematological indicators, particularly the eosinophil-to-neutrophil ratio (EN ratio), to enhance diagnostic accuracy. Experimental results demonstrate the excellent performance of the BREATHE method in ECRS diagnosis. The combination of E-M scoring with the EN ratio yielded the best results, significantly improving diagnostic accuracy. By employing an ensemble method using logistic regression and random forest models, we further enhanced predictive performance. The main innovations of this study include: proposing an automated sinus segmentation algorithm that does not rely on large-scale datasets and developing an effective CT image screening technique. These innovations provide new tools and insights for early diagnosis and precise treatment of ECRS, with the potential to significantly improve patients' quality of life.