在双能光子计数CT中对HU到RED校准的有效原子数进行深度学习估计
1Medical Information Research Section, Electronics and Telecommunications Research Institute, Daejeon, Republic of Korea.
Scientific reports
|December 23, 2025
概括
一种新的深度学习模型准确地预测有效原子数 (EAN) 并改善双能CT (DECT) 中的相对电子密度 (RED) 估计. 这种进步增强了材料的分化,并支持精确的剂量计算.
科学领域:
- 医学物理 医学物理
- 医疗成像中的人工智能
背景情况:
- 精确的材料表征在双能CT (DECT) 中至关重要,用于诸如放射治疗规划等应用.
- 估计材料属性的常规方法,如有效原子数 (EAN) 和相对电子密度 (RED),都有局限性.
研究的目的:
- 开发和评估一个深度学习 (DL) 框架,用于精确的 EAN 预测,以改善 DECT.中 RED 估计.
- 为了比较DL方法的性能与传统的拉瑟福和静电测量方法.
主要方法:
- 一个修改后的U-Net深度学习模型被训练在来自光子计数探测器CT的合成光谱数据上.
- 该模型直接估计了EAN,然后使用基于物理的模型将其转换为RED.
- 对八种材料的性能进行了评估,使用平均绝对误差 (MAE) 和余量,将DL与传统方法进行比较.
主要成果:
- 与拉瑟福 (1.59%) 和静态测量 (1.54%) 方法相比,DL模型实现了EAN (0.08%) 的显著较低的MAE.
- 基于DL的RED估计得出MAE为0.62%,残留量最小.
- 与传统方法 (R2 ≈ 0.9995) 相比,DL方法在 ΔHU-RED 校准曲线中显示出略高的线性 (R2 = 0.9998).
结论:
- 深度学习框架显著提高了DECT幻影研究中EAN和RED估计的准确性.
- 通过DL方法实现的增强HU-RED线性提高了材料差异化能力.
- 这种DL方法有可能在未来的临床DECT应用中进行更精确的剂量计算.
更多相关视频
06:28Visualization of Low-Level Gamma Radiation Sources Using a Low-Cost, High-Sensitivity, Omnidirectional Compton Camera
Published on: January 30, 2020
13.2K
06:20Irradiator Commissioning and Dosimetry for Assessment of LQ α and β Parameters, Radiation Dosing Schema, and in vivo Dose Deposition
Published on: March 11, 2021
7.6K
相关概念视频
Atomic Absorption Spectroscopy: Radiation and Light Sources
1.1K
Atomic absorption spectroscopy (AAS) relies on the Beer-Lambert law, which requires that the radiation source emits a narrow range of wavelengths to match the absorption characteristics of the analyte atom. The primary criteria for choosing an appropriate radiation source in AAS is to provide a precise and intense emission at specific wavelengths that will allow accurate detection of the analyte.
Two common narrow-range 'line' sources used in AAS are hollow-cathode lamps (HCLs) and...
Two common narrow-range 'line' sources used in AAS are hollow-cathode lamps (HCLs) and...
1.1K
Atomic Emission Spectroscopy: Overview
3.4K
Atomic emission spectroscopy (AES) is an analytical technique used to determine the elemental composition of a sample by analyzing the light emitted from excited atoms. In AES, atoms in a sample are excited to higher energy levels by thermal energy from high-temperature sources, such as plasma, arcs, or sparks. When these excited atoms return to lower energy states, they emit light at specific wavelengths characteristic of each element. The resulting atomic emission spectrum, which consists of...
3.4K
Atomic Radii and Effective Nuclear Charge
61.4K
The elements in groups of the periodic table exhibit similar chemical behavior. This similarity occurs because the members of a group have the same number and distribution of electrons in their valence shells.
61.4K
