等中心
成像体模
光子计数
核医学
图像分辨率
像素
计算机科学
人工智能
基准标记
生物医学工程
材料科学
探测器
模式识别(心理学)
医学
电信
作者
Shobhit Sharma,Debashish Pal,Ehsan Abadi,Thomas J. Sauer,Paul Segars,Jiang Hsieh,Ehsan Samei
标识
DOI:10.1016/j.acra.2022.06.018
摘要
Deep silicon-based photon-counting CT (Si-PCCT) is an emerging detector technology that provides improved spatial resolution by virtue of its reduced pixel sizes. This article reports the outcomes of the first simulation study evaluating the impact of this advantage over energy-integrating CT (ECT) for estimation of morphological radiomics features in lung lesions.A dynamic nutrient-access-based stochastic model was utilized to generate three distinct morphologies for lung lesions. The lesions were inserted into the lung parenchyma of an anthropomorphic phantom (XCAT - 50th percentile BMI) at 50, 70, and 90 mm from isocenter. The phantom was virtually imaged with an imaging simulator (DukeSim) modeling a Si-PCCT and a conventional ECT system using varying imaging conditions (dose, reconstruction kernel, and pixel size). The imaged lesions were segmented using a commercial segmentation tool (AutoContour, Advantage Workstation Server 3.2, GE Healthcare) followed by extraction of morphological radiomics features using an open-source radiomics package (pyradiomics). The estimation errors for both systems were computed as percent differences from corresponding feature values estimated for the ground-truth lesions.Compared to ECT, the mean estimation error was lower for Si-PCCT (independent features: 35.9% vs. 54.0%, all features: 54.5% vs. 68.1%) with statistically significant reductions in errors for 8/14 features. For both systems, the estimation accuracy was minimally affected by dose and distance from the isocenter while reconstruction kernel and pixel size were observed to have a relatively stronger effect.For all lesions and imaging conditions considered, Si-PCCT exhibited improved estimation accuracy for morphological radiomics features over a conventional ECT system, demonstrating the potential of this technology for improved quantitative imaging.
科研通智能强力驱动
Strongly Powered by AbleSci AI