本研究使用飛利浦的能譜CT（Philips IQon Spectral CT）成像參數的影像，包括：一般CT、單能CT（40keV）、碘密度、含碘去水的碘密度和有效原子序影像，定量分析椎骨轉移病灶，探討能譜CT的各種成像參數下影像是否具備診斷分辨骨轉移。篩選30位病人具椎骨轉移，分割骨轉移病灶，使用深度學習分割出正常椎骨。透過ROC分析法，找出最佳閾值以鑑別骨轉移病灶。椎骨轉移病灶和正常椎骨在能譜CT上有明顯差異，適合使用閥值來區分，能譜CT各參數分析的AUC皆大於0.9。分類結果顯示有效原子序影像的AUC最佳（0.942），碘密度影像則較差（0.905）。有效原子序影像具有最高的敏感度（89.19%），含碘去水的碘密度影像有最高的特異性（93.32%）。能譜CT的成像參數皆能有效鑑別椎骨轉移。

This study uses Philips IQon Spectral CT imaging parameters including conventional CT, monoenergetic CT (40 keV), iodine density, iodine-no-water, and Z effective images to conduct a quantitative analysis on bone metastasis lesions. The feasibility of using various spectral CT imaging parameters to diagnose bone metastasis was discussed. Thirty patients with vertebral bone metastasis were selected and manually segmented. Deep learning was used to segment the normal vertebrae. The optimal threshold for identifying bone metastasis lesions was found by ROC analysis. Bone metastasis lesions and normal vertebrae showed significant differences in spectral CT, which were suitable for using a threshold to distinguish. The AUC of each parameter of the spectral CT images was greater than 0.9. The classification results showed that the effective Z image had the best AUC (0.942), while the image of the iodine density had the worst performance (0.905). The effective Z image also had the highest sensitivity (89.19%), and the iodine-no-water image had the highest specificity (93.32%). Different spectral CT imaging parameters can effectively identify bone metastases.