Korean Journal of Radiology

  • 제22권3호
  • /
  • Pages.344-353
  • /
  • 2021
  • /
  • 1229-6929(pISSN)
  • /
  • 2005-8330(eISSN)
대한영상의학회 (The Korean Society of Radiology)
권호 표지
DOI QR코드

DOI QR Code

Development of a Malignancy Potential Binary Prediction Model Based on Deep Learning for the Mitotic Count of Local Primary Gastrointestinal Stromal Tumors

  • Jiejin Yang (Department of Radiology, Peking University First Hospital, Peking University) ;
  • Zeyang Chen (Department of General Surgery, Peking University First Hospital, Peking University) ;
  • Weipeng Liu (Beijing Smart Tree Medical Technology Co. Ltd.) ;
  • Xiangpeng Wang (Beijing Smart Tree Medical Technology Co. Ltd.) ;
  • Shuai Ma (Department of Radiology, Peking University First Hospital, Peking University) ;
  • Feifei Jin (Department of Biostatistics, Peking University First Hospital) ;
  • Xiaoying Wang (Department of Radiology, Peking University First Hospital, Peking University)
  • 투고 : 2019.11.11
  • 심사 : 2020.06.15
  • 발행 : 2021.03.01
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

Objective: The mitotic count of gastrointestinal stromal tumors (GIST) is closely associated with the risk of planting and metastasis. The purpose of this study was to develop a predictive model for the mitotic index of local primary GIST, based on deep learning algorithm. Materials and Methods: Abdominal contrast-enhanced CT images of 148 pathologically confirmed GIST cases were retrospectively collected for the development of a deep learning classification algorithm. The areas of GIST masses on the CT images were retrospectively labelled by an experienced radiologist. The postoperative pathological mitotic count was considered as the gold standard (high mitotic count, > 5/50 high-power fields [HPFs]; low mitotic count, ≤ 5/50 HPFs). A binary classification model was trained on the basis of the VGG16 convolutional neural network, using the CT images with the training set (n = 108), validation set (n = 20), and the test set (n = 20). The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated at both, the image level and the patient level. The receiver operating characteristic curves were generated on the basis of the model prediction results and the area under curves (AUCs) were calculated. The risk categories of the tumors were predicted according to the Armed Forces Institute of Pathology criteria. Results: At the image level, the classification prediction results of the mitotic counts in the test cohort were as follows: sensitivity 85.7% (95% confidence interval [CI]: 0.834-0.877), specificity 67.5% (95% CI: 0.636-0.712), PPV 82.1% (95% CI: 0.797-0.843), NPV 73.0% (95% CI: 0.691-0.766), and AUC 0.771 (95% CI: 0.750-0.791). At the patient level, the classification prediction results in the test cohort were as follows: sensitivity 90.0% (95% CI: 0.541-0.995), specificity 70.0% (95% CI: 0.354-0.919), PPV 75.0% (95% CI: 0.428-0.933), NPV 87.5% (95% CI: 0.467-0.993), and AUC 0.800 (95% CI: 0.563-0.943). Conclusion: We developed and preliminarily verified the GIST mitotic count binary prediction model, based on the VGG convolutional neural network. The model displayed a good predictive performance.

키워드

참고문헌

  1. Zhao X, Yue C. Gastrointestinal stromal tumor. J Gastrointest Oncol 2012;3:189-208
  2. Miettinen M, Lasota J. Gastrointestinal stromal tumors: review on morphology, molecular pathology, prognosis, and differential diagnosis. Arch Pathol Lab Med 2006;130:1466-1478 https://doi.org/10.5858/2006-130-1466-GSTROM
  3. Rammohan A, Sathyanesan J, Rajendran K, Pitchaimuthu A, Perumal SK, Srinivasan U, et al. A gist of gastrointestinal stromal tumors: a review. World J Gastrointest Oncol 2013;5:102-112 https://doi.org/10.4253/wjge.v5.i3.102
  4. Joensuu H. Risk stratification of patients diagnosed with gastrointestinal stromal tumor. Hum Pathol 2008;39:1411-1419 https://doi.org/10.1016/j.humpath.2008.06.025
  5. Miettinen M, Lasota J. Gastrointestinal stromal tumors: pathology and prognosis at different sites. Semin Diagn Pathol 2006;23:70-83 https://doi.org/10.1053/j.semdp.2006.09.001
  6. Poveda A, Garcia Del Muro X, Lopez-Guerrero JA, Cubedo R, Martinez V, Romero I, et al. GEIS guidelines for gastrointestinal sarcomas (GIST). Cancer Treat Rev 2017;55:107-119 https://doi.org/10.1016/j.ctrv.2016.11.011
  7. von Mehren M, Randall RL, Benjamin RS, Boles S, Bui MM, Ganjoo KN, et al. Soft tissue sarcoma, version 2.2018, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw 2018;16:536-563 https://doi.org/10.6004/jnccn.2018.0025
  8. Sepe PS, Brugge WR. A guide for the diagnosis and management of gastrointestinal stromal cell tumors. Nat Rev Gastroenterol Hepatol 2009;6:363-371 https://doi.org/10.1038/nrgastro.2009.43
  9. Demetri GD, von Mehren M, Antonescu CR, DeMatteo RP, Ganjoo KN, Maki RG, et al. NCCN task force report: update on the management of patients with gastrointestinal stromal tumors. J Natl Compr Canc Netw 2010;8 Suppl 2:S1-41. https://doi.org/10.6004/jnccn.2010.0116
  10. Ha CY, Shah R, Chen J, Azar RR, Edmundowicz SA, Early DS. Diagnosis and management of GI stromal tumors by EUSFNA: a survey of opinions and practices of endosonographers. Gastrointest Endosc 2009;69:1039-1044.e1. https://doi.org/10.1016/j.gie.2008.07.041
  11. Vernuccio F, Taibbi A, Picone D, LA Grutta L, Midiri M, Lagalla R, et al. Imaging of Gastrointestinal Stromal Tumors: From Diagnosis to Evaluation of Therapeutic Response. Anticancer Res 2016;36:2639-2648
  12. Lee MW, Kim SH, Kim YJ, Lee JM, Lee JY, Park EA, et al. Gastrointestinal stromal tumor of the stomach: preliminary results of preoperative evaluation with CT gastrography. Abdom Imaging 2008;33:255-261 https://doi.org/10.1007/s00261-007-9253-x
  13. Greenfield D, Wilson S. Artificial intelligence in medicine: applications, implications, and limitations. Availabe at: http://sitn.hms.harvard.edu/flash/2019/artificialintelligence-in-medicine-applications-implications-andlimitations/. Accessed September 24, 2019
  14. Beam AL, Kohane IS. Translating artificial intelligence into clinical care. JAMA 2016;316:2368-2369 https://doi.org/10.1001/jama.2016.17217
  15. Aerts HJWL. Data science in radiology: a path forward. Clin Cancer Res 2018;24:532-534
  16. Seif G. Deep learning vs classical machine learning. Available at: https://towardsdatascience.com/deep-learning-vsclassical-machine-learning-9a42c6d48aa. Accessed September 25, 2019
  17. Vial A, Stirling D, Field M, Ros M, Ritz C, Carolan M, et al. The role of deep learning and radiomic feature extraction in cancer-specific predictive modelling: a review. Transl Cancer Res 2018;7:803-816 https://doi.org/10.21037/tcr.2018.05.02
  18. Wang S, Shi J, Ye Z, Dong D, Yu D, Zhou M, et al. Predicting EGFR mutation status in lung adenocarcinoma on computed tomography image using deep learning. Eur Respir J 2019;53:1800986
  19. Jung H, Kim B, Lee I, Lee J, Kang J. Classification of lung nodules in CT scans using three-dimensional deep convolutional neural networks with a checkpoint ensemble method. BMC Med Imaging 2018;18:48.
  20. Song Q, Zhao L, Luo X, Dou X. Using deep learning for classification of lung nodules on computed tomography images. J Healthc Eng 2017;2017:8314740
  21. Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. Proceedings of the ICLR;2015 May 7-9, San Diego, CA, USA
  22. Dematteo RP, Gold JS, Saran L, Gonen M, Liau KH, Maki RG, et al. Tumor mitotic rate, size, and location independently predict recurrence after resection of primary gastrointestinal stromal tumor (GIST). Cancer 2008;112:608-615 https://doi.org/10.1002/cncr.23199
  23. Miettinen M, Lasota J, Sobin LH. Gastrointestinal stromal tumors of the stomach in children and young adults: a clinicopathologic, immunohistochemical, and molecular genetic study of 44 cases with long-term follow-up and review of the literature. Am J Surg Pathol 2005;29:1373-1381 https://doi.org/10.1097/01.pas.0000172190.79552.8b
  24. Ishikawa T, Kanda T, Kameyama H, Wakai T. Neoadjuvant therapy for gastrointestinal stromal tumor. Transl Gastroenterol Hepatol 2018;3:3
  25. von Mehren M. The role of adjuvant and neoadjuvant therapy in gastrointestinal stromal tumors. Curr Opin Oncol 2008;20:428-432 https://doi.org/10.1097/CCO.0b013e328302ed82
  26. Doyon C, Sideris L, Leblanc G, Leclerc YE, Boudreau D, Dube P. Prolonged therapy with imatinib mesylate before surgery for advanced gastrointestinal stromal tumor results of a phase II trial. Int J Surg Oncol 2012;2012:761576
  27. Wang C, Li H, Jiaerken Y, Huang P, Sun L, Dong F, et al. Building CT radiomics-based models for preoperatively predicting malignant potential and mitotic count of gastrointestinal stromal tumors. Transl Oncol 2019;12:1229-1236 https://doi.org/10.1016/j.tranon.2019.06.005
  28. Pelandre GL, Djahjah MC, Gasparetto EL, Nacif MS, Marchiori E, Mello EL. Tomographic findings of gastric gastrointestinal stromal tumor and correlation with the mitotic index. Arq Gastroenterol 2013;50:244-250 https://doi.org/10.1590/S0004-28032013000400002
  29. Iannicelli E, Carbonetti F, Federici GF, Martini I, Caterino S, Pilozzi E, et al. Evaluation of the relationships between computed tomography features, Pathological findings, and prognostic risk assessment in gastrointestinal stromal tumors. J Comput Assist Tomogr 2017;41:271-278 https://doi.org/10.1097/RCT.0000000000000499
  30. Zhou C, Duan X, Zhang X, Hu H, Wang D, Shen J. Predictive features of CT for risk stratifications in patients with primary gastrointestinal stromal tumour. Eur Radiol 2016;26:3086-3093 https://doi.org/10.1007/s00330-015-4172-7
  31. Li H, Ren G, Cai R, Chen J, Wu X, Zhao J. A correlation research of Ki67 index, CT features, and risk stratification in gastrointestinal stromal tumor. Cancer Med 2018;7:4467-4474 https://doi.org/10.1002/cam4.1737
  32. Zhuo T, Li X, Zhou H. Combining radiomics and CNNs to classify benign and malignant GIST. Proceedings of the 2018 International Conference on Network, Communication, Computer Engineering (NCCE 2018); 2018 May 26-27; Chongqing, China: Atlantis Press; 2018
  33. Ning Z, Luo J, Li Y, Han S, Feng Q, Xu Y, et al. Pattern classification for gastrointestinal stromal tumors by integration of radiomics and deep convolutional features. IEEE J Biomed Health Inform 2019;23:1181-1191 https://doi.org/10.1109/JBHI.2018.2841992
  34. Chen T, Ning Z, Xu L, Feng X, Han S, Roth HR, et al. Radiomics nomogram for predicting the malignant potential of gastrointestinal stromal tumours preoperatively. Eur Radiol 2019;29:1074-1082 https://doi.org/10.1007/s00330-018-5629-2
  35. Feng C, Lu F, Shen Y, Li A, Yu H, Tang H, et al. Tumor heterogeneity in gastrointestinal stromal tumors of the small bowel: volumetric CT texture analysis as a potential biomarker for risk stratification. Cancer Imaging 2018;18:46
  36. Maier A, Syben C, Lasser T, Riess C. A gentle introduction to deep learning in medical image processing. Z Med Phys 2019;29:86-101 https://doi.org/10.1016/j.zemedi.2018.12.003
  37. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature 2015;521:436-444
  38. Watabe T, Tatsumi M, Watabe H, Isohashi K, Kato H, Yanagawa M, et al. Intratumoral heterogeneity of F-18 FDG uptake differentiates between gastrointestinal stromal tumors and abdominal malignant lymphomas on PET/CT. Ann Nucl Med 2012;26:222-227 https://doi.org/10.1007/s12149-011-0562-3
  39. Miettinen M, Makhlouf H, Sobin LH, Lasota J. Gastrointestinal stromal tumors of the jejunum and ileum: a clinicopathologic, immunohistochemical, and molecular genetic study of 906 cases before imatinib with long-term follow-up. Am J Surg Pathol 2006;30:477-489 https://doi.org/10.1097/00000478-200604000-00008
  40. Parab TM, DeRogatis MJ, Boaz AM, Grasso SA, Issack PS, Duarte DA, et al. Gastrointestinal stromal tumors: a comprehensive review. J Gastrointest Oncol 2019;10:144-154 https://doi.org/10.21037/jgo.2018.08.20
  41. Pinaikul S, Woodtichartpreecha P, Kanngurn S, Leelakiatpaiboon S. Gastrointestinal stromal tumor (GIST): computed tomographic features and correlation of CT findings with histologic grade. J Med Assoc Thai 2014;97:1189-1198