Summary

实验鼠模型中骨代谢的早期检测机器学习算法

Published: August 16, 2020
doi:

Summary

该协议旨在训练机器学习算法,以结合从磁共振成像 (MRI) 和正电子发射断层扫描 (PET/CT) 中乳腺癌骨转移的大鼠模型中得出的成像参数,以检测早期转移性疾病并预测宏位值的后续进展。

Abstract

机器学习 (ML) 算法允许将不同要素集成到模型中,以执行精度超过其成分的分类或回归任务。该协议描述了ML算法的发展,用于预测大鼠模型中乳腺癌骨骼微量酶的生长,然后通过标准成像方法观察到任何异常。这种算法可以促进早期转移性疾病(即微虫病)的检测,这种疾病在分期检查期间经常遗漏。

应用的转移模型是站点特定的,这意味着大鼠只在右后腿发展转移。该模型的肿瘤接受率为60%~80%,在诱导30天后,在动物子集中,磁共振成像(MRI)和正电子发射断层扫描/计算机断层扫描(PET/CT)中,宏位酶变得可见,而第二部分动物则没有肿瘤生长。

从早期获得的图像检查开始,本协议描述了通过MRI检测到的指示组织血管化特征的提取、PET/CT检测到的葡萄糖代谢,以及随后确定预测宏观代谢疾病最相关特征的特征。然后,这些功能被输入到模型平均神经网络(avNNet)中,将动物分为两组:一组会发展转移,另一组不会发展任何肿瘤。该协议还描述了标准诊断参数的计算,例如总体精度、灵敏度、特异性、负/阳性预测值、可能性比以及接收器操作特性的制定。建议的协议的优点是它的灵活性,因为它可以很容易地适应,以训练大量的不同的ML算法与无限数量的功能的可调组合。此外,它可以用来分析不同的问题在肿瘤学,感染和炎症。

Introduction

该协议的目的是将MRI和PET/CT的多个功能成像参数集成到模型平均神经网络(avNNet)ML算法中。该算法预测乳腺癌骨转移大鼠模型中宏位值的生长,此时骨骼内的宏观变化尚不可见。

在巨肉生长之前,骨髓入侵传播的肿瘤细胞,俗称微米静病,1,2。1这种最初的入侵可以被认为是转移性疾病的早期步骤,但通常在常规分期检查3,4,期间错过。虽然目前可用的成像模式不能检测骨髓微创时单独使用,结合成像参数产生血管化和代谢活动的信息已被证明是效果更好5。这种互补的好处是通过将不同的成像参数组合成一个avNet,这是一个ML算法。这种 avNNet 允许在存在任何可见转移之前可靠地预测骨宏酶的形成。因此,将成像生物标志物集成到avNet中可以作为骨髓微创和早期转移性疾病的代理参数。

为了开发这个方案,在裸体大鼠中使用了先前描述的乳腺癌骨转移模型,使用了6,7,8。,7,8这种模式的优点是它的站点特异性,这意味着动物在右后腿只发展骨质转移。然而,这种方法的肿瘤接受率为60%~80%,因此相当多的动物在研究期间不会发展出任何转移。使用成像方式(如 MRI 和 PET/CT),从注射后第 30 天 (PI) 开始可检测到转移的存在。在较早的时间点(例如,10 PI)成像不区分动物将发展转移性疾病和那些不会(图1)。

avNNet 接受过关于第 10 天 PI 上采集的功能成像参数的训练,如以下协议所述,可可靠地预测或排除宏位值在未来 3 周内的生长。神经网络将人工节点组合在不同的层内。在研究方案中,骨髓供血和代谢活动的功能成像参数代表底层,恶性肿瘤的预测代表顶层。其他中间层包含连接到顶部层和底部层的隐藏节点。在网络训练期间更新不同节点之间的连接强度,以高精度执行各自的分类任务9。通过平均多个模型的输出,可以进一步提高这种神经网络的精度,从而产生一个 avNNet10

Protocol

所有护理和实验程序都按照国家和区域动物保护立法进行,所有动物程序都得到德国弗兰科尼亚州政府的批准(参考号55.2 DMS-2532-2228)。 1. 裸体大鼠右后腿乳腺癌骨转移的诱导 注:关于裸体大鼠乳腺癌骨转移诱导的详细说明已在其他地方发表,6,8。,8下面列出了最相关的步骤。 培养MDA-MB-231人类乳腺癌…

Representative Results

大鼠从MDA-MB-231乳腺癌细胞的手术和注射中迅速恢复,然后于第10天和30日接受M-和PET/CT成像(图1)。图 2A中对大鼠右近性头骨进行了具有代表性的DCE分析。通过选择”导出”按钮并选择”DCEraw.txt”作为文件名来保存 DCE 原始度量。 动态参数、AUC、PE 和洗涤的后续计算在 RStudio 中用相应的脚本执行。DCE 测量值的输出必须保存?…

Discussion

ML 算法是强大的工具,用于将多个预测功能集成到组合模型中,并在单独使用时获得超过其单独成分的精度。尽管如此,实际结果取决于几个关键步骤。首先,使用的ML算法是一个关键因素,因为不同的ML算法产生不同的结果。此协议中使用的算法是 avNNet,但其他有前途的算法包括极端梯度提升21 或 随机林。RStudio 的 caret 包20 提供了大量不同的算法(目前是 >17…

Disclosures

The authors have nothing to disclose.

Acknowledgements

这项工作得到了德国研究基金会(DFG,合作研究中心CRC 1181,子项目Z02的支持;优先方案 +Bone,项目 BA 4027/10-1 和 BO 3811),包括对扫描设备(INST 410/77-1 FUGG 和 INST 410/93-1 FUGG)和新兴领域倡议 (EFI) “大 Thera” 的弗里德里希- 亚历山大大学埃尔兰根-纽伦堡提供的额外支持。

Materials

Binocular Operating Microscope Leica NA
ClinScan MR System Bruker NA
DICOM Viewer Horos NA www.horosproject.org
Excel: Spreadsheet Microsoft NA
FCS Sigma F2442-500ML
Gadovist Bayer-Schering NA
Inveon PET/CT Siemens NA
Inveon Research Workplace Software Siemens Healthcare GmbH NA
IVIS Spectrum PerkinElmer NA
MDA-MB-231 human breast cancer cells American Type Culture Collection N/A
Open-source data visualization, machine learning and data mining toolkit. Orange3, University of Ljubljana NA https://orange.biolab.si/
RPMI-1640 Invitrogen/ThermoFisher 11875093
Trypsin Sigma 9002-07-7
Vevo 3100 VisualSonics NA

References

  1. D’Oronzo, S., Brown, J., Coleman, R. The role of biomarkers in the management of bone-homing malignancies. Journal of Bone Oncology. 9, 1-9 (2017).
  2. Ellmann, S., Beck, M., Kuwert, T., Uder, M., Bäuerle, T. Multimodal imaging of bone metastases: From preclinical to clinical applications. Journal of Orthopaedic Translation. 3 (4), 166-177 (2015).
  3. Braun, S., Pantel, K. Clinical significance of occult metastatic cells in bone marrow of breast cancer patients. The Oncologist. 6 (2), 125-132 (2001).
  4. Braun, S., Rosenberg, R., Thorban, S., Harbeck, N. Implications of occult metastatic cells for systemic cancer treatment in patients with breast or gastrointestinal cancer. Seminars in surgical oncology. 20 (4), 334-346 (2001).
  5. Ellmann, S., et al. Prediction of early metastatic disease in experimental breast cancer bone metastasis by combining PET/CT and MRI parameters to a Model-Averaged Neural Network. Bone. 120, 254-261 (2018).
  6. Bäuerle, T., Komljenovic, D., Berger, M. R., Semmler, W. Multi-modal imaging of angiogenesis in a nude rat model of breast cancer bone metastasis using magnetic resonance imaging, volumetric computed tomography and ultrasound. Journal of Visualized Experiments. (66), e4178 (2012).
  7. Merz, M., Komljenovic, D., Semmler, W., Bäuerle, T. Quantitative contrast-enhanced ultrasound for imaging antiangiogenic treatment response in experimental osteolytic breast cancer bone metastases. Investigative Radiology. 47 (7), 422-429 (2012).
  8. Bäuerle, T., et al. Characterization of a rat model with site-specific bone metastasis induced by MDA-MB-231 breast cancer cells and its application to the effects of an antibody against bone sialoprotein. International Journal of Cancer. 115 (2), 177-186 (2005).
  9. Patel, J., Goyal, R. Applications of Artificial Neural Networks in Medical Science. Current Clinical Pharmacology. 2 (3), 217-226 (2008).
  10. Naftaly, U., Intrator, N., Horn, D. Optimal ensemble averaging of neural networks. Network: Computation in Neural Systems. 8 (3), 283-296 (1997).
  11. Bäuerle, T., Merz, M., Komljenovic, D., Zwick, S., Semmler, W. Drug-induced vessel remodeling in bone metastases as assessed by dynamic contrast enhanced magnetic resonance imaging and vessel size imaging: A longitudinal in vivo study. Clinical Cancer Research. 16 (12), 3215-3225 (2010).
  12. Cheng, C., et al. Evaluation of treatment response of cilengitide in an experimental model of breast cancer bone metastasis using dynamic PET with 18F-FDG. Hellenic Journal of Nuclear Medicine. 14 (1), 15-20 (2011).
  13. Marturano-Kruik, A., et al. Human bone perivascular niche-on-a-chip for studying metastatic colonization. Proceedings of the National Academy of Sciences of the United States of America. 115 (6), 1256-1261 (2018).
  14. Sonntag, E., et al. In vivo proof-of-concept for two experimental antiviral drugs, both directed to cellular targets, using a murine cytomegalovirus model. Antiviral Research. 161, 63-69 (2019).
  15. . Horos – Free DICOM Medical Image Viewer | Open-Source Available from: https://www.horosproject.org/ (2015)
  16. . RStudio Team RStudio: Inteegrated Development for R Available from: https://rstudio.com (2015)
  17. Demšar, J., et al. Orange: Data Mining Toolbox in Python. Journal of Machine Learning Research. 14, 2349-2353 (2013).
  18. Saeys, Y., Inza, I., Larrañaga, P. A review of feature selection techniques in bioinformatics. Bioinformatics. 23 (19), 2507-2517 (2007).
  19. . CRAN – Package caret Available from: https://cran.r-project.org/web/packages/caret/index.html (2016)
  20. . CRAN: Package xgboost – Extreme Gradient Boosting Available from: https://cran.r-project.org/web/packages/xgboost/ (2019)
  21. Fernández-Delgado, M., Cernadas, E., Barro, S., Amorim, D., Fernández-Delgado, A. Do we Need Hundreds of Classifiers to Solve Real World Classification Problems. Journal of Machine Learning Research. 15, 3133-3181 (2014).
  22. Hira, Z. M., Gillies, D. F. A Review of Feature Selection and Feature Extraction Methods Applied on Microarray Data. Advances in Bioinformatics. 2015, 198363 (2015).
  23. Sánchez-Maroño, N., Alonso-Betanzos, A., Tombilla-Sanromán, M. Filter methods for feature selection – A comparative study. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 4881, 178-187 (2007).
  24. Cawley, G. C., Talbot, N. L. C. C. Fast exact leave-one-out cross-validation of sparse least-squares support vector machines. Neural Network. 17 (10), 1467-1475 (2004).
  25. Forghani, R., et al. Radiomics and Artificial Intelligence for Biomarker and Prediction Model Development in Oncology. Computational and Structural Biotechnology Journal. 17, 995-1008 (2019).
  26. Jaffe, C. C. Measures of response: RECIST, WHO, and new alternatives. Journal of Clinical Oncology Official Journal of the American Society of Clinical Oncology. 24 (20), 3245-3251 (2006).
  27. Lambin, P., et al. Radiomics: Extracting more information from medical images using advanced feature analysis. European Journal of Cancer. 48 (4), 441-446 (2012).
  28. Gillies, R. J., Kinahan, P. E., Hricak, H. Radiomics: Images are more than pictures, they are data. Radiology. 278 (2), 563-577 (2016).
  29. Nioche, C., et al. Lifex: A freeware for radiomic feature calculation in multimodality imaging to accelerate advances in the characterization of tumor heterogeneity. 암 연구학. 78 (16), 4786-4789 (2018).
  30. Ellmann, S., et al. Application of machine learning algorithms for multiparametric MRI-based evaluation of murine colitis. PLOS ONE. 13 (10), 0206576 (2018).
check_url/kr/61235?article_type=t

Play Video

Cite This Article
Ellmann, S., Seyler, L., Gillmann, C., Popp, V., Treutlein, C., Bozec, A., Uder, M., Bäuerle, T. Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model. J. Vis. Exp. (162), e61235, doi:10.3791/61235 (2020).

View Video