基于幻影和临床图像数据的3D FDG PET分割的多位点质量和变异性分析(英文)

2017-02-08 09:37:58 admin 21

Multi-site quality and variability analysis of 3D FDG PET segmentations based on phantom and clinical image data.

Med Phys. 2017 Feb

Beichel RR1,2Smith BJ3Bauer C1Ulrich EJ1,4Ahmadvand P5Budzevich MM6Gillies RJ6Goldgof D7Grkovski M8Hamarneh G5Huang Q9Kinahan PE10Laymon CM11,12Mountz JM12Muzi JP10Muzi M10Nehmeh S13Oborski MJ11Tan Y9Zhao B9Sunderland JJ14Buatti JM15.

Author information

    1.Department of Electrical and Computer Engineering, The University of Iowa, Iowa City, IA, USA.

    2.Department of Internal Medicine, The University of Iowa, Iowa City, IA, USA.

    3.Department of Biostatistics, The University of Iowa, Iowa City, IA, USA.

    4.Department of Biomedical Engineering, The University of Iowa, Iowa City, IA, USA.

    5.School of Computing Science, Simon Fraser University, Burnaby, Canada.

    6.H Lee Moffitt Cancer Center, Tampa, FL, USA.

    7.Department of Computer Science and Engineering, University of South Florida, Tampa, FL, USA.

    8.Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, USA.

    9.Department of Radiology, Columbia University Medical Center, New York, NY, USA.

    10.Department of Radiology, University of Washington Medical Center, Seattle, WA, USA.

    11.Department of Bioengineering, University of Pittsburgh, Pittsburgh, PA, USA

    12.Department of Radiology, University of Pittsburgh, Pittsburgh, PA, USA.

    13.National Center for Cancer Care and Research, Doha, Qatar.

    14.Department of Radiology, The University of Iowa, Iowa City, IA, USA.

    15.Department of Radiation Oncology, The University of Iowa, Iowa City, IA, USA.

Abstract

PURPOSE


Radiomics utilizes a large number of image-derived features for quantifying tumor characteristics that can in turn be correlated with response and prognosis. Unfortunately, extraction and analysis of such image-based features is subject to measurement variability and bias. The challenge for radiomics is particularly acute in Positron Emission Tomography (PET) where limited resolution, a high noise component related to the limited stochastic nature of the raw data, and the wide variety of reconstruction options confound quantitative feature metrics. Extracted feature quality is also affected by tumor segmentation methods used to define regions over which to calculate features, making it challenging to produce consistent radiomics analysis results across multiple institutions that use different segmentation algorithms in their PET image analysis. Understanding each element contributing to these inconsistencies in quantitative image feature and metric generation is paramount for ultimate utilization of these methods in multi-institutional trials and clinical oncology decision making.

METHODS


To assess segmentation quality and consistency at the multi-institutional level, we conducted a study of seven institutional members of the National Cancer Institute Quantitative Imaging Network. For the study, members were asked to segment a common set of phantom PET scans acquired over a range of imaging conditions as well as a second set of head and neck cancer (HNC) PET scans. Segmentations were generated at each institution using their preferred approach. In addition, participants were asked to repeat segmentations with a time interval between initial and repeat segmentation. This procedure resulted in overall 806 phantom insert and 641 lesion segmentations. Subsequently, the volume was computed from the segmentations and compared to the corresponding reference volume by means of statistical analysis.

RESULTS


On the two test sets (phantom and HNC PET scans), the performance of the seven segmentation approaches was as follows. On the phantom test set, the mean relative volume errors ranged from 29.9 to 87.8% of the ground truth reference volumes, and the repeat difference for each institution ranged between -36.4 to 39.9%. On the HNC test set, the mean relative volume error ranged between -50.5 to 701.5%, and the repeat difference for each institution ranged between -37.7 to 31.5%. In addition, performance measures per phantominsert/lesion size categories are given in the paper. On phantom data, regression analysis resulted in coefficient of variation (CV) components of 42.5% for scanners, 26.8% for institutional approaches, 21.1% for repeated segmentations, 14.3% for relative contrasts, 5.3% for count statistics (acquisition times), and 0.0% for repeated scans. Analysis showed that the CV components for approaches and repeated segmentations were significantly larger on the HNC test set with increases by 112.7% and 102.4%, respectively.

CONCLUSION


Analysis results underline the importance of PET scanner reconstruction harmonization and imaging protocol standardization for quantification of lesion volumes. In addition, to enable a distributed multi-site analysis of FDG PET images, harmonization of analysisapproaches and operator training in combination with highly automated segmentation methods seems to be advisable. Future work will focus on quantifying the impact of segmentation variation on radiomics system performance.

KEYWORDS


FDG PET; head and neck cancer; multi-site performance analysis; phantom; radiomics; segmentation


慧影医疗科技(北京)有限公司

地点:北京市海淀区中关村东升科技园B2-C103

电话:400-890-9020

邮箱:radcloud@huiyihuiying.com

关闭
图片
图片
  • 人工智能诊断云平台