现代仪器与医疗

基于隐式神经表示的能谱CT重建与物质分解方法研究

PDF下载

夏颖贤 ^a^,^b , 周浩 ^a^,^b , 邓一凡 ^a^,^b , 王志磊 ^a^,^b , 陈志强 ^a^,^b , 高河伟 ^a^,^b^,^*

现代仪器与医疗 | CT关键技术专栏 2026,32(2): 1-6

收起

现代仪器与医疗 |CT关键技术专栏 2026 , 32 (2) : 1 -6

基于隐式神经表示的能谱CT重建与物质分解方法研究

全屏

夏颖贤^a^,^b, 周浩^a^,^b, 邓一凡^a^,^b, 王志磊^a^,^b, 陈志强^a^,^b, 高河伟^a^,^b^,^*

作者信息

^a.清华大学　工程物理系，北京　100084

^b.清华大学　粒子技术与辐射成像教育部重点实验室，北京　100084

通讯作者:

高河伟，男，博士研究生，教授，研究方向：CT成像技术及应用

作者简介:

夏颖贤，男，博士研究生，助理研究员，研究方向：CT成像技术及应用

Research on spectral CT reconstruction and material decomposition method based on implicit neural representation

Yingxian Xia^a^,^b, Hao Zhou^a^,^b, Yifan Deng^a^,^b, Zhilei Wang^a^,^b, Zhiqiang Chen^a^,^b, Hewei Gao^a^,^b^,^*

Affiliations

^a.Department of Engineering Physics, Tsinghua University, Beijing 100084, China

^b.Key Laboratory of Particle & Radiation Imaging, Ministry of Education, Tsinghua University, Beijing 100084, China

出版时间: 2026-04-10 doi: 10.11876/mimt202602001

文章导航

摘要

收起

目的

解决能谱CT传统重建方法依赖离散体素网格、计算复杂度高，且在低剂量与欠采样条件下重建质量下降、物质分解精度不足的技术难题。

方法

提出一种基于隐式神经表示（implicit neural representation，INR）的能谱CT重建与物质分解方法。该方法采用多分辨率哈希编码将空间坐标映射为高维特征，并通过轻量级多层感知机（multi-layer perceptron，MLP）解码输出基物质密度，将重建过程转化为最小化投影一致性损失的非线性优化问题。

结果

基于快速千伏切换仿真数据与双层平板探测器实际数据完成性能验证，与传统的投影域物质分解结合滤波反投影（filter back projection，FBP）的两步法相比，该方法有效提升了图像的信噪比与物质分解精度，并在稀疏视图条件下展现出更强的鲁棒性。

结论

本方法可为新一代能谱CT技术的低剂量临床应用提供新的算法路径与理论依据。

关键词

隐式神经表示 / 能谱CT / 物质分解 / 图像重建

Abstract

收起

Objective

To address the technical challenges of traditional spectral CT reconstruction methods, including reliance on discrete voxel grids, high computational complexity, degraded reconstruction quality, and insufficient material decomposition accuracy under low-dose and under-sampling conditions.

Methods

A spectral CT reconstruction and material decomposition method based on implicit neural representation (INR) is proposed. This method adopts multi-resolution hash encoding to map spatial coordinates into high-dimensional features, decodes base material densities via a lightweight multi-layer perceptron (MLP), and converts the reconstruction process into a nonlinear optimization problem minimizing projection consistency loss.

Results

Performance verification was conducted based on simulation data from fast kV switching and actual data from a dual-layer flat-panel detector. Compared with the traditional two-step method combining projection-domain material decomposition and filtered back projection (FBP), the proposed method effectively improves the image signal-to-noise ratio and material decomposition accuracy, demonstrating stronger robustness under sparse-view conditions.

Conclusion

The present invention can provide a new algorithmic path and theoretical basis for the low-dose clinical application of next-generation spectral CT technology.

Key words

Implicit neural representation / Spectral CT / Material decomposition / Image reconstruction

引用本文

夏颖贤, 周浩, 邓一凡, 王志磊, 陈志强, 高河伟. 基于隐式神经表示的能谱CT重建与物质分解方法研究. 现代仪器与医疗, 2026 , 32 (2) : 1 -6 . DOI: 10.11876/mimt202602001

Yingxian Xia, Hao Zhou, Yifan Deng, Zhilei Wang, Zhiqiang Chen, Hewei Gao. Research on spectral CT reconstruction and material decomposition method based on implicit neural representation[J]. Modern Instruments & Medical Treatment, 2026 , 32 (2) : 1 -6 . DOI: 10.11876/mimt202602001

基金

收起

国家重点研发计划“诊疗装备与生物医用材料”重点专项(2024YFC2417500)

参考文献

收起

文献

收起

参考文献引证文献

排序方式：

[1]

Kalender W A. Computed tomography: fundamentals, system technology, image quality, applications[M]. 3rd ed. Hoboken: John Wiley & Sons, 2011.

[2]

Peters T M. Computed tomography: fundamentals, system technology, image quality and applications[J]. Physics in Medicine & Biology, 2001, 46(5): 1593.

[3]

Seeram E. Computed tomography: physical principles, clinical applications, and quality control[M]. 4th ed. Amsterdam: Elsevier, 2015.

[4]

Marin D, Boll D T, Mileto A, et al. State of the art: dualenergy CT of the abdomen[J]. Radiology, 2014, 271(2): 327-342.

[5]

Schilham A, van Hamersvelt R W, de Jong P A, et al. Material analysis through X-ray attenuation decomposition in spectral computed tomography[J]. Scientific Reports, 2025, 15: 27714.

[6]

Kim B, Shim H, Baek J. A streak artifact reduction algorithm in sparse-view CT using a self-supervised neural representation[J]. Medical Physics, 2022, 49(12): 7497-7515.

[7]

Mory C, Sixou B, Si-Mohamed S, et al. Comparison of five one-step reconstruction algorithms for spectral CT[J]. Physics in Medicine & Biology, 2018, 63(23): 235001.

[8]

Chen X, Xia W, Yang Z, et al. SOUL-net: a sparse and low-rank unrolling network for spectral CT image reconstruction[J]. IEEE Transactions on Neural Networks and Learning Systems, 2024, 35(12): 18620-18634.

[9]

Molaei A, Aminimehr A, Tavakoli A, et al. Implicit neural representation in medical imaging: a comparative survey[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2023: 2381-2391.

[10]

Zhu Y, Liu Y, Zhang Y, et al. Implicit neural representation for medical image reconstruction[J]. Physics in Medicine & Biology, 2025, 70(12): 12TR01.

[11]

Shen L, Pauly J, Xing L. NeRP: implicit neural representation learning with prior embedding for sparsely sampled image reconstruction[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022, 35(1): 770-782.

[12]

Zhang Y, Shao H C, Pan T, et al. Dynamic cone-beam CT reconstruction using spatial and temporal implicit neural representation learning (STINR)[J]. Physics in Medicine & Biology, 2023, 68(4): 045005.

[13]

Lee J, Baek J. Iterative reconstruction for limited-angle CT using implicit neural representation[J]. Physics in Medicine & Biology, 2024, 69(10): 105008.

[14]

Song B, Shen L, Xing L. Piner: prior-informed implicit neural representation learning for test-time adaptation in sparse-view CT reconstruction[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2023: 1928-1938.

[15]

Najaf M, Ongie G. Accelerated optimization of implicit neural representations for CT reconstruction[C]//Proceedings of IEEE 22nd International Symposium on Biomedical Imaging (ISBI). Piscataway: IEEE, 2025: 1-5.

[16]

Wang Y, Liang N, Wang S, et al. Implicit neural prior-guided diffusion for spectral CT reconstruction[J]. Medical Physics, 2025, 52(7): e17946.

[17]

Müller T, Evans A, Schied C, et al. Instant neural graphics primitives with a multiresolution hash encoding[J]. ACM Transactions on Graphics (TOG), 2022, 41(4): 1-15.

[18]

Mildenhall B, Srinivasan P P, Tancik M, et al. Nerf: representing scenes as neural radiance fields for view synthesis[J]. Communications of the ACM, 2022, 65(1): 99-106.

[19]

Poludniowski G, Landry G, Deblois F, et al. SpekCalc: a program to calculate photon spectra from tungsten anode x-ray tubes[J]. Physics in Medicine & Biology, 2009, 54(19): N433.

[20]

Wang Z, Zhou H, Gu S, et al. Dual-energy head cone-beam CT using a dual-layer flat-panel detector: hybrid material decomposition and a feasibility study[J]. Medical Physics, 2023, 50(11): 6762-6778.

2026年第32卷第2期

PDF下载

引用本文

BibTeX

文章信息

doi: 10.11876/mimt202602001

首发时间：2026-10-23
出版时间：2026-04-10

补充材料

相关文章

文章信息

作者

出版历史

基金

国家重点研发计划“诊疗装备与生物医用材料”重点专项(2024YFC2417500)

作者信息

^a.清华大学　工程物理系，北京　100084

^b.清华大学　粒子技术与辐射成像教育部重点实验室，北京　100084

通讯作者:

高河伟，男，博士研究生，教授，研究方向：CT成像技术及应用

参考文献

分享链接

https://castjournals.cast.org.cn/joweb/xdyqyyl/CN/10.11876/mimt202602001

分享至

全文二维码

扫描看全文

引用本文

BibTeX

本文的引用情况

2种不同金属材料的力学参数

科 Family	属数 Number of genus	种数 Number of species	占总种数比例 Percentage of total species (%)	属 Genus	种数 Number of species	占总种数比例 Percentage of total species (%)
鹅膏菌科Amanitaceae	2	11	5.26	鹅膏菌属 Amanita	10	4.78
小菇科 Mycenaceae	2	12	5.74	丝盖伞属 Inocybe	5	2.39
多孔菌科 Polyporaceae	8	14	6.70	蜡蘑属 Laccaria	5	2.39
红菇科 Russulaceae	3	23	11.00	小皮伞属 Marasmius	6	2.87
				小菇属 Mycena	11	5.26
				光柄菇属 Pluteus	5	2.39
				红菇属 Russula	17	8.13
				栓菌属 Trametes	5	2.39

关闭全屏

BibTeX
EndNote
RefWorks
TxT