胸部電腦斷層掃描是肺部疾病診斷的最重要的成像方式之一，肺部影像分割在電腦輔助分析和診斷過程中扮演至關重要的作用。由於深度學習模型在醫學影像解剖結構的語義分割方面的準確性已經達到了人類水平，我們提議使用經過訓練的深度學習模型來預測胸部電腦斷層切片中的肺部、肺炎感染區域和心臟位置。使用以肺炎診斷導向形式的3D影像視覺化透視人體結構繪製技術對語義分割結果進行總結和感染區域之視覺化。本次研究使用了五筆公開的胸腔電腦斷層資料集來訓練SE-ResNeXt50-FPN深度學習模型來訓練分割出胸腔電腦斷層中的肺部、感染區域和心臟位置。透過重疊率IoU與組內相關系數(ICC)的統計運算得出，模型在分割的結果上與專家所標記有很高的吻合度。我們將分割結果來生成肺炎導向的電腦輔助診斷地圖，該地圖是由3D的解剖結構的投影位置、肺總體積、和肺感染體積訊息組成，可用作為檢查胸部電腦斷層疑似肺部病變的位置指南，並可合併為診斷治療急性肺部疾病的快速分流演算法。;Computer tomography (CT) scanning of the chest is one of the most important imaging modalities available for pulmonary disease diagnosis. Lung segmentation plays a crucial step in the pipeline of computer-aided analysis and diagnosis. As deep learning models have achieved human-level accuracy in semantic segmentation of anatomical structures, we propose using trained deep learning models to segment lung, heart, and the infectious areas of pneumonia in chest CT scan slices. In this study, five publicly available thoracic CT data sets were used in training and testing a deep learning semantic segmentation model, SE-ResNeXt50-FPN, for the segmentation of lungs, infected regions, and heart locations in chest CT. The statistical analyses of intersect-over-union (IoU) overlapping scores, Bland-Altman plot, and intraclass correlation coefficients (ICC) demonstrated that the segmentation results of the model were highly consistent with the annotations marked by human experts. The semantic segmentation results were summarized and visualized using volume rendering technology to generate a pneumonia-oriented interpretation map consisting of both location and volume information that can be used as location guidance for inspecting suspected pulmonary lesions of chest CT and can possibly be combined into a rapid triage algorithm for treating acute pulmonary diseases.
