基于胸部CT的人工智能骨密度测量系统机会性筛查骨质疏松症的可行性研究

崔璨; 温庆祥; 刘妮; 李民; 李俊秋; 刘军莲; 霍健伟

doi:10.15953/j.ctta.2024.289

基于胸部CT的人工智能骨密度测量系统机会性筛查骨质疏松症的可行性研究

崔璨^a,,
温庆祥^b,
刘妮^a,
李民^a,
李俊秋^a,
刘军莲^a, ,,
霍健伟^a, ,

a.
首都医科大学附属北京中医医院放射科；
b.
首都医科大学附属北京中医医院核医学科，北京 100010

详细信息

作者简介:
崔璨，女，硕士，放射科主治医师，主要从事骨密度测量方面的研究，E-mail：540545201@qq.com

通讯作者:
刘军莲✉，女，副主任技师，主要从事骨密度测量方面的研究，E-mail：13501192985@163.com

霍健伟✉，男，主任医师，主要从事影像诊断工作，E-mail:huojianwei@bjzhongyi.com。

中图分类号: R 814
计量
- 文章访问数: 30
- HTML全文浏览量: 16
- PDF下载量: 11
出版历程
- 收稿日期: 2024-12-04
- 修回日期: 2025-01-25
- 录用日期: 2025-02-05
- 网络出版日期: 2025-03-12

Feasibility of opportunistic osteoporosis screening using an artificial intelligence-based bone density measurement on chest CT scans

a.
Beijing Hospital of Traditional Chinese Medicine, Capital Medical University, Beijing 100010, China Department of Radiology
b.
Beijing Hospital of Traditional Chinese Medicine, Capital Medical University, Beijing 100010, China Department of Nuclear Medicine

摘要

摘要:
探讨基于胸部CT的人工智能（AI）骨密度测量系统机会性筛查骨质疏松症的可行性。回顾性分析于我科2023年8月至2024年7月同时行双光能X线吸收测定法（DXA）和胸部CT的462名患者的资料，其中绝经后的女性317例、50岁以上的男性145例。比较两种方法测量骨密度之间的差异；以DXA测量的T值为参考标准，分析基于胸部CT的AI系统与DXA测量结果的一致性和相关性。绝经后女性和50岁以上男性的身高、体重、DXA T值及AI BMD有统计学差异；AI测量的BMD与DXA T值之间的相关系数为0.767；二者的κ值为0.697；AI诊断骨质疏松的ROC曲线下面积为0.941（95%CI 0.914~0.968），敏感性85.71%，特异性93.84%。AI骨密度测量系统与DXA测定骨密度具有高度相关性及良好一致性，可以帮助机会性筛查骨质疏松症。
- CT /
- 人工智能 /
- 骨密度 /
- 骨质疏松
Abstract:
This study explores the feasibility of opportunistic osteoporosis screening using an artificial intelligence (AI)-based bone mineral density (BMD) measurement system on chest computed tomography (CT) scans. A retrospective analysis was conducted on 462 patients who underwent both dual-energy X-ray absorptiometry (DXA) and chest CT in our department between August 2023 and July 2024. The cohort included 317 postmenopausal women and 145 men aged > 50 years. BMD measurements from the AI system and DXA were compared. Using the T-value measured by DXA as the reference standard, the consistency and correlation between AI-based and DXA-measured BMD were analyzed. Significant differences in height, weight, DXA T-value, and AI-derived BMD were observed between men aged > 50 years and postmenopausal women. The AI-derived BMD showed a correlation coefficient of 0.767 with DXA T-values and a κ value of 0.697. The area under the ROC curve for AI-based diagnosis of osteoporosis was 0.941（95% CI 0.914–0.968), with a sensitivity of 85.71% and a specificity of 93.84%. The AI-based BMD measurement system demonstrates strong correlation and good agreement with DXA, supporting its feasibility for opportunistic osteoporosis screening.
- CT /
- Artificial intelligence /
- Bone density /
- osteoporosis

HTML全文

图 1 AI测量BMD与DXA测量T值的相关性散点图

Figure 1. Scatter plot of correlation between BMD measured by AI and T value measured by DXA

下载: 全尺寸图片幻灯片

图 2 AI诊断骨质疏松症的ROC曲线

Figure 2. ROC curve of AI diagnosis of osteoporosis

下载: 全尺寸图片幻灯片

表 1 462名受试者基本资料

Table 1 Demographic data from 462 subjects

变量	全部（n=462）	女性（n=317）	男性（n=145）	P值
年龄（岁）	64.94±8.547	64.99±8.221	64.81±9.249	0.837
身高（cm）	162.98±7.278	159.35±4.855	170.90±5.065	＜0.001
体重（kg）	65.91±11.065	62.52±9.609	73.34±10.412	＜0.001
BMI（kg/m²）	24.75±3.385	24.60±3.492	25.08±3.125	0.160
DXA T值	−1.79±0.855	−1.98±0.839	−1.36±0.726	＜0.001
AI BMD（mg/cm³）	111.33±33.522	106.66±32.886	121.53±32.732	＜0.001

下载: 导出CSV

表 2 DXA和AI诊断类型的分布

Table 2 Distributions of DXA and AI diagnostic types

		AI
		骨量正常	骨量减少	骨质疏松	总计
DXA	骨量正常	67（14.5%）	23（5.0%）^a	1（0.2%）^b	91（19.7%）
	骨量减少	21（4.5%）^a	225（48.7%）	20（4.3%）^a	266（57.6%）
	骨质疏松	0（0%）^b	16（3.5%）^a	89（19.3%）	105（22.7%）
	总计	88（19.0%）	264（57.1%）	110（23.8%）	462（100%）
注：黑体字为一致；a为轻度不一致：一种方法显示骨密度正常，而另一种则显示骨量减少或一种方法显示骨量减少，而另一种则显示骨质疏松；b为重度不一致：一种方法显示骨密度正常，而另一种则显示骨质疏松。

下载: 导出CSV

参考文献(20)

[1]	LIU F, ZHU H, MA J, et al. Performance of iCare quantitative computed tomography in bone mineral density assessment of the hip and vertebral bodies in European spine phantom[J]. Journal of Orthopaedic Surgery and Research, 2023, 18(1): 777. DOI: 10.1186/s13018-023-04174-w.
[2]	CUI Z, MENG X, FENG H, et al. Estimation and projection about the standardized prevalence of osteoporosis in mainland China[J]. Archives of Osteoporosis, 2019, 15(1): 2. DOI: 10.1007/s11657-019-0670-6.
[3]	SI L, WINZENBERG T M, JIANG Q, et al. Projection of osteoporosis-related fractures and costs in China: 2010-2050[J]. Osteoporosis International, 2015, 26(7): 1929-1937. DOI: 10.1007/s00198-015-3093-2.
[4]	DE MARGERIE-MELLON C, CHASSAGNON G. Artificial intelligence: A critical review of applications for lung nodule and lung cancer[J]. Diagnostic and Interventional Imaging, 2023, 104(1): 11-17. DOI: 10.1016/j.diii.2022.11.007.
[5]	孙安, 樊荣荣, 孙瑶, 等. CT重组算法对低剂量胸部CT筛查冠状动脉钙化积分准确性影响研究[J]. 临床放射学杂志, 2022, 41(5): 881-885. DOI: 10.13437/j.cnki.jcr.2022.05.025. SUN A, FAN R, SUN Y, et al. The effect of CT reconstruction kernel on the accuracy of low-dose chest CT screening for coronary artery calcification scores[J]. Journal of Clinical Radiology, 2022, 41(5): 881-885. DOI: 10.13437/j.cnki.jcr.2022.05.025.
[6]	赵宇, 张晓岚, 郑超, 等. 基于低剂量胸部CT深度学习模型自动测量骨密度研究[J]. 放射学实践, 2024, 39(2): 262-266. DOI: 10.13609/j.cnki.1000-0313.2024.02.019. ZHAO Y, ZHANG X L, ZHENG C, et al. Bone densitometry measurement based on low-dose chest CT with deep learning model[J]. Radiol Practice, 2024, 39(2): 262-266. DOI: 10.13609/j.cnki.1000-0313.2024.02.019.
[7]	赵君禄, 刘斋, 韩康, 等. 基于定量CT的机会性骨质疏松诊断: 2种定量CT骨密度测量软件临床应用比较分析[J]. 河北医科大学学报, 2024, 45(1): 17-23. DOI: 10.3969/j.issn.1007-3205.2024.01.005. ZHAO J L, LIU Z, HAN K, et al. Computed tomography—based opportunistic osteoporosis diagnosis: A comparison of clinical applications of two quantitative CT softwares[J]. Journal of Hebei Medical University, 2024, 45(1): 17-23. DOI: 10.3969/j.issn.1007-3205.2024.01.005.
[8]	SIRIS E S, ADLER R, BILEZIKIAN J, et al. The clinical diagnosis of osteoporosis: a position statement from the National Bone Health Alliance Working Group[J]. Osteoporosis International, 2014, 25(5): 1439-1443. DOI: 10.1007/s00198-014-2655-z.
[9]	BUDOFF M J, HAMIRANI Y S, GAO Y L, et al. Measurement of thoracic bone mineral density with quantitative CT[J]. Radiology, 2010, 257(2): 434-440. DOI: 10.1148/radiol.10100132.
[10]	KENDRICK J, FRANCIS R J, HASSAN G M, et al. Prognostic utility of RECIP 1.0 with manual and AI-based segmentations in biochemically recurrent prostate cancer from [(68)Ga]Ga-PSMA-11 PET images[J]. European Journal of Nuclear Medicine and Molecular Imaging, 2023, 50(13): 4077-4086. DOI: 10.1007/s00259-023-06382-2.
[11]	GERETY E L, HOPPER M A, BEARCROFT P W. The reliability of measuring the density of the L1 vertebral body on CT imaging as a predictor of bone mineral density[J]. Clinical Radiology, 2017, 72(2): 177-179. DOI: 10.1016/j.crad.2016.09.022.
[12]	YAN L, WANG X, YU T, et al. Characteristics of the gut microbiota and serum metabolites in postmenopausal women with reduced bone mineral density[J]. Frontiers in Cellular and Infection Microbiology, 2024, 14: 1367325. DOI: 10.3389/fcimb.2024.1367325.
[13]	WU X, ZHANG M. Effects of androgen and progestin on the proliferation and differentiation of osteoblasts[J]. Experimental and Therapeutic Medicine, 2018, 16(6): 4722-4728. DOI: 10.3892/etm.2018.6772.
[14]	KIM Y W, KIM J H, YOON S H, et al. Vertebral bone attenuation on low-dose chest CT: Quantitative volumetric analysis for bone fragility assessment[J]. Osteoporosis International, 2017, 28(1): 329-338. DOI: 10.1007/s00198-016-3724-2.
[15]	SAVAGE R H, van ASSEN M, MARTIN S S, et al. Utilizing Artificial Intelligence to Determine Bone Mineral Density Via Chest Computed Tomography[J]. Journal of Thoracic Imaging, 2020, 35 Suppl 1: S35-S39. DOI: 10.1097/RTI.0000000000000484.
[16]	LIN W, HE C, XIE F, et al. Quantitative CT screening improved lumbar BMD evaluation in older patients compared to dual-energy X-ray absorptiometry[J]. Bmc Geriatrics, 2023, 23(1): 231. DOI: 10.1186/s12877-023-03963-6.
[17]	王盟盟, 张磊, 周凤云, 等. 双层光谱CT、QCT及DXA在骨质疏松诊断中的精确性与效能比较[J]. CT理论与应用研究, 2024, 33(6): 717-724. DOI: 10.15953/j.ctta.2024.086. WANG M M, ZHANG L, ZHOU F Y, et al. Comparative Study of the Accuracies and Efficiencies of Dual-layer Spectral CT, QCT, and DXA for Osteoporosis Diagnosis[J]. CT Theory and Applications, 2024, 33(6): 717-724. DOI: 10.15953/j.ctta.2024.086.
[18]	EBBESEN E N, THOMSEN J S, BECK-NIELSEN H, et al. Vertebral bone density evaluated by dual-energy X-ray absorptiometry and quantitative computed tomography in vitro[J]. Bone, 1998, 23(3): 283-290. DOI: 10.1016/s8756-3282(98)00091-x.
[19]	WONG M, PAPA A, LANG T, et al. Validation of thoracic quantitative computed tomography as a method to measure bone mineral density[J]. Calcified Tissue International, 2005, 76(1): 7-10. DOI: 10.1007/s00223-004-0020-5.
[20]	张羽, 张宗军, 刘许慧, 等. 胸椎定量CT和腰椎双能X线吸收检测仪对绝经后女性骨质疏松症的诊断差异[J]. 放射学实践, 2022, 37(10): 1205-1210. DOI: 10.13609/j.cnki.1000-0313.2022.10.003. ZHANG Y, ZHANG Z J, LIU X H, et al. Diagnosis difference of thoracic spine quantitative CT and lumbar spine DXA for osteoporosis in postmenopausal women[J]. Radiol Practice, 2022, 37(10): 1205-1210. DOI: 10.13609/j.cnki.1000-0313.2022.10.003.