目的： 探讨不同扫描条件下，基于光子计数CT物质分离技术衍生的脂肪分数（CT-derived fat fraction，CT-FF）与磁共振成像质子密度脂肪分数（magnetic resonance imaging proton density fat fraction， MRI-PDFF）间的一致性，以期建立适用于中国人群的肝脏CT脂肪含量的测定方法。方法： 2023年9月至2024年2月期间，上海交通大学医学院附属瑞金医院前瞻性招募了383位健康志愿者（MRI-PDFF < 5%者176例，MRI-PDFF≥5%者207例），根据CT管电压（120 kVp/140 kVp）和辐射剂量（标准剂量/低剂量）不同，将其随机分配至不同光子计数CT扫描方案的4组［①120 kVp+标准剂量（n = 123）；②120 kVp+低剂量（n = 71）；③140 kVp+标准剂量（n = 120）；④140 kVp+低剂量（n = 69）］。所有受试者均接受光子计数CT肝脏扫描和MRI检查，并测量肝脏MRI-PDFF值作为肝脏脂肪含量测定的金标准。在纳入人群（n = 383）的标准剂量组（n = 243）内，随机挑选管电压120 kVp组（n = 123）和140 kVp组（n = 120）中各50人，组成测试队列（n = 100），剩余受试者作为验证队列（n = 283）。在测试队列的MRI-PDFF<5%的志愿者（n = 66）中，分别在120 kVp组（n = 33）和140 kVp组（n = 33）各随机选取20人，组成阈值调整队列（n =40），测量肝脏和腹壁皮下脂肪组织在高、低能量箱下的平均CT值，作为物质分离阈值。在测试队列中，观察采用对比调整前（机器提供的参考阈值）阈值，测得CT-FF值与MRI-PDFF值的相关性和一致性，同时观察采用调整队列获得的阈值测得CF-FF与MRI-PDFF值间的相关性和一致性。结果： 基于阈值调整队列数据，120 kVp下，肝脏组织在低、高能量箱的平均CT值分别为65 HU和70 HU，140 kVp下低、高能量箱平均CT值亦是65 HU、70 HU；脂肪组织在120 kVp低、高能量箱的平均CT值分别为−127 HU 和−96 HU，在140 kVp低、高能量箱的平均CT值分别为−125 HU和−92 HU，以上作为物质分离密度阈值。在测试队列中，阈值调整后CT-FF与MRI-PDFF的相关性（r，0.98比0.77）、一致性（ICC，0.980比0.770；r²，0.96比0.60）较前（基于机器提供的参考阈值测得的结果）明显提升，平均差值显著缩小（−0.7%比−18.1%）。在验证队列整组和不同的管电压及辐射剂量亚组中，CT-FF值与MRI-PDFF值的相关性和一致性都极好（r = 0.99， P < 0.001， r² = 0.98， ICC = 0.99），平均差值均不大于−0.7%。 结论： 本研究基于中国人肝脏组织特性，优化光子计数CT物质分离算法的密度阈值，首次建立了适用于国人的脂肪定量校正标准，显著提升测量准确性，有望为无创、精准定量肝脏脂肪含量提供新手段。

Objective To investigate the consistency between CT-derived fat fraction (CT-FF) based on photoncounting CT material decomposition under different scanning conditions and magnetic resonance imaging proton density fat fraction (MRI-PDFF), thereby developing a CT-based method for liver fat quantification suitable for the Chinese population. Methods From September 2023 to February 2024, a total of 383 healthy volunteers were prospectively recruited at Ruijin Hospital, Shanghai Jiao Tong University School of Medicine (176 with MRI-PDFF < 5% and 207 with MRI-PDFF≥ 5%), and randomly assigned to four photon-counting CT scanning groups based on tube voltage (120 kVp/140 kVp) and radiation dose (standard dose/low dose) [Different CT scanning schemes:① 120 kVp+standard dose (n = 123); ② 120 kVp+low dose (n = 71); ③ 140 kVp+standard dose (n = 120); ④ 140 kVp+low dose (n = 69)]. All subjects underwent photoncounting CT liver scanning and MRI examinations, with liver MRI-PDFF used as the reference standard for liver fat quantification. From the standard-dose group (n = 243), this study randomly selected 50 individuals each from the 120 kVp group (n = 123) and 140 kVp group (n = 120) to form a test cohort (n = 100), and the remaining subjects were assigned to the validation cohort (n = 283). Among volunteers with MRI-PDFF < 5% (n = 66) in the test cohort, this study randomly selected 20 individuals each from the 120 kVp group (n = 33) and 140 kVp group (n = 33) to form a threshold adjustment cohort (n = 40). The average CT values of liver and subcutaneous abdominal fat tissues were measured under low and high energy bins to serve as the adjusted thresholds for material decomposition. In the test cohort, the correlation and consistency between CT-FF and MRI-PDFF values obtained using thresholds before (the threshold value provided by the machine itself) and after adjustment (threshold obtained from the adjustment queue) were compared. The performance of the adjusted threshold in measuring liver fat content was evaluated in the validation cohort, as well as the consistency across subgroups with different scanning protocols. Results Based on data from the threshold adjustment cohort, the average CT values of liver tissue at 120 kVp were 65 HU and 70 HU in the low and high energy bins, and so were at 140 kVp. For fat tissue, the average CT values in the low and high energy bins were −127 HU and −96 HU at 120 kVp, and −125 HU and −92 HU at 140 kVp, which were used as the density thresholds for material decomposition. In the test cohort, after threshold adjustment, the correlation (r, 0.98 vs. 0.77), consistency (ICC, 0.980 vs. 0.770; r², 0.96 vs. 0.60), and mean difference (−0.7% vs. −18.1%) between CT-FF and MRI-PDFF values were significantly improved. In the entire validation cohort and subgroups with different tube voltages and radiation doses, CT-FF and MRI-PDFF showed excellent correlation and consistency (r = 0.99, P < 0.001, r² = 0.98, ICC = 0.99), with mean differences not exceeding −0.7%. Conclusion Based on the liver tissue characteristics of the Chinese population, this study optimizes the density thresholds of the photon-counting CT material decomposition algorithm, and develops a fat quantification correction standard applicable to Chinese individuals for the first time, significantly improving measurement accuracy. This method may provide a new non-invasive and precise approach for liver fat quantification.