Skip to main content

Advertisement

SpringerLink
Log in

Validation of a deep learning segmentation algorithm to quantify the skeletal muscle index and sarcopenia in metastatic renal carcinoma

  • Musculoskeletal
  • Published:
European Radiology Aims and scope Submit manuscript

Abstract

Objectives

To validate a deep learning (DL) algorithm for measurement of skeletal muscular index (SMI) and prediction of overall survival in oncology populations.

Methods

A retrospective single-center observational study included patients with metastatic renal cell carcinoma between 2007 and 2019. A set of 37 patients was used for technical validation of the algorithm, comparing manual vs DL-based evaluations. Segmentations were compared using mean Dice similarity coefficient (DSC), SMI using concordance correlation coefficient (CCC) and Bland-Altman plots. Overall survivals (OS) were compared using log-rank (Kaplan-Meier) and Mann-Whitney tests. Generalizability of the prognostic value was tested in an independent validation population (N = 87).

Results

Differences between two manual segmentations (DSC = 0.91, CCC = 0.98 for areas) or manual vs. automated segmentation (DSC = 0.90, CCC = 0.98 for areas, CCC = 0.97 for SMI) had the same order of magnitude. Bland-Altman plots showed a mean difference of −3.33 cm2 [95%CI: −15.98, 9.1] between two manual segmentations, and −3.28 cm2 [95% CI: −14.77, 8.21] for manual vs. automated segmentations. With each method, 20/37 (56%) patients were classified as sarcopenic. Sarcopenic vs. non-sarcopenic groups had statistically different survival curves with median OS of 6.0 vs. 12.5 (p = 0.008) and 6.0 vs. 13.9 (p = 0.014) months respectively for manual and DL methods. In the independent validation population, sarcopenic patients according to DL had a lower OS (10.7 vs. 17.3 months, p = 0.033).

Conclusion

A DL algorithm allowed accurate estimation of SMI compared to manual reference standard. The DL-calculated SMI demonstrated a prognostic value in terms of OS.

Key Points

• A deep learning algorithm allows accurate estimation of skeletal muscle index compared to a manual reference standard with a concordance correlation coefficient of 0.97.

• Sarcopenic patients according to SMI thresholds after segmentation by the deep learning algorithm had statistically significantly lower overall survival compared to non-sarcopenic patients.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4

Similar content being viewed by others

Abbreviations

BMI:

Body mass index

CCC:

Concordance correlation coefficient

CI:

Confidence interval

CT:

Computed tomography

DL:

Deep learning

DSC:

Dice similarity coefficient

ECOG:

Eastern Cooperative Oncology Group

FCN:

Fully convolutional network

mRCC:

Metastatic renal cell carcinoma

OS:

Overall survival

RCC:

Renal cell carcinoma

RMSE:

Root mean square error

SD:

Standard deviation

SMI:

Skeletal muscular index

References

  1. Cruz-Jentoft AJ, Bahat G, Bauer J et al (2019) Sarcopenia: revised European consensus on definition and diagnosis. Age Ageing 48:16–31. https://doi.org/10.1093/ageing/afy169

    Article  PubMed  Google Scholar 

  2. Shachar SS, Williams GR, Muss HB, Nishijima TF (2016)) Prognostic value of sarcopenia in adults with solid tumours: a meta-analysis and systematic review. Eur J Cancer 57:58–67. https://doi.org/10.1016/j.ejca.2015.12.030

  3. Auclin E, Bourillon C, De Maio E et al (2017) Prediction of everolimus toxicity and prognostic value of skeletal muscle index in patients with metastatic renal cell carcinoma. Clin Genitourin Cancer 15:350–355. https://doi.org/10.1016/j.clgc.2017.01.022

    Article  PubMed  Google Scholar 

  4. O’Connor JPB, Aboagye EO, Adams JE et al (2017) Imaging biomarker roadmap for cancer studies. Nat Rev Clin Oncol 14:169–186. https://doi.org/10.1038/nrclinonc.2016.162

    Article  CAS  PubMed  Google Scholar 

  5. Blanc-Durand P, Schiratti J-B, Schutte K et al (2020) Abdominal musculature segmentation and surface prediction from CT using deep learning for sarcopenia assessment. Diagn Interv Imaging. https://doi.org/10.1016/j.diii.2020.04.011

  6. Lassau N, Bousaid I, Chouzenoux E et al (2020) Three artificial intelligence data challenges based on CT and MRI. Diagn Interv Imaging 101:783–788. https://doi.org/10.1016/j.diii.2020.03.006

    Article  CAS  PubMed  Google Scholar 

  7. Dice LR (1945) Measures of the amount of ecologic association between species. Ecology 26:297–302. https://doi.org/10.2307/1932409

    Article  Google Scholar 

  8. Nie Z, Xu J, Zhang S (2020) Analysis on DeepLabV3+ performance for automatic steel defects detection. ArXiv200404822 Cs

  9. Nagarajan P, Warnell G, Stone P (2019) Deterministic implementations for reproducibility in deep reinforcement learning. ArXiv180905676 Cs

  10. Goodpaster BH, Thaete FL, Kelley DE (2000) Composition of skeletal muscle evaluated with computed tomography. Ann N Y Acad Sci 904:18–24. https://doi.org/10.1111/j.1749-6632.2000.tb06416.x

    Article  CAS  PubMed  Google Scholar 

  11. Mourtzakis M, Prado CMM, Lieffers JR et al (2008) A practical and precise approach to quantification of body composition in cancer patients using computed tomography images acquired during routine care. Appl Physiol Nutr Metab 33:997–1006. https://doi.org/10.1139/H08-075

  12. Fearon K, Strasser F, Anker SD et al (2011) Definition and classification of cancer cachexia: an international consensus. Lancet Oncol 12:489–495. https://doi.org/10.1016/S1470-2045(10)70218-7

    Article  PubMed  Google Scholar 

  13. Yushkevich PA, Piven J, Hazlett HC et al (2006) User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage 31:1116–1128. https://doi.org/10.1016/j.neuroimage.2006.01.015

  14. Martin L, Birdsell L, Macdonald N et al (2013) Cancer cachexia in the age of obesity: skeletal muscle depletion is a powerful prognostic factor, independent of body mass index. J Clin Oncol Off J Am Soc Clin Oncol 31:1539–1547. https://doi.org/10.1200/JCO.2012.45.2722

    Article  Google Scholar 

  15. Zijdenbos AP, Dawant BM, Margolin RA, Palmer AC (1994) Morphometric analysis of white matter lesions in MR images: method and validation. IEEE Trans Med Imaging 13:716–724. https://doi.org/10.1109/42.363096

    Article  CAS  PubMed  Google Scholar 

  16. Grenier B, Dubreuil M, Journois D (2000) Comparaison de deux méthodes de mesure d’une même grandeur : méthode de Bland et Altman. Ann Fr Anesth Reanim 19:128–135. https://doi.org/10.1016/S0750-7658(00)00109-X

  17. Chai T, Draxler RR (2014) Root mean square error (RMSE) or mean absolute error (MAE)?– Arguments against avoiding RMSE in the literature. Geosci Model Dev 7:1247–1250. https://doi.org/10.5194/gmd-7-1247-2014

    Article  Google Scholar 

  18. Brodersen KH, Ong CS, Stephan KE, Buhmann JM (2010) The balanced accuracy and its posterior distribution. In: 2010 20th International Conference on Pattern Recognition. pp 3121–3124

  19. Mouracade P (2017) Key concepts of survival analysis: checking appropriateness. Prog Urol Soc Francaise Urol 27:331–333. https://doi.org/10.1016/j.purol.2017.03.012

  20. Burns JE, Yao J, Chalhoub D et al (2020) A machine learning algorithm to estimate sarcopenia on abdominal CT. Acad Radiol 27:311–320. https://doi.org/10.1016/j.acra.2019.03.011

    Article  PubMed  Google Scholar 

  21. Hu P, Huo Y, Kong D, et al (2018) Automated characterization of body composition and frailty with clinically acquired CT. Comput Methods Clin Appl Musculoskelet Imaging 5th Int Workshop MSKI 2017 Held Conjunction MICCAI 2017 Quebec City QC Can Sept 10 2017 Revis Sel Pap MSKI Work 10734:25–35. https://doi.org/10.1007/978-3-319-74113-0_3

  22. Castiglione J, Somasundaram E, Gilligan LA et al (2021) Automated segmentation of abdominal skeletal muscle on pediatric CT scans using deep learning. Radiol Artif Intell 3:e200130. https://doi.org/10.1148/ryai.2021200130

    Article  PubMed  PubMed Central  Google Scholar 

  23. Shahedi M, Cool DW, Romagnoli C et al (2014) Spatially varying accuracy and reproducibility of prostate segmentation in magnetic resonance images using manual and semiautomated methods. Med Phys 41:113503. https://doi.org/10.1118/1.4899182

    Article  PubMed  Google Scholar 

  24. Hermoye L, Laamari-Azjal I, Cao Z et al (2005) Liver segmentation in living liver transplant donors: comparison of semiautomatic and manual methods. Radiology 234:171–178. https://doi.org/10.1148/radiol.2341031801

    Article  PubMed  Google Scholar 

  25. El-Bana S, Al-Kabbany A, Sharkas M (2020) A two-stage framework for automated malignant pulmonary nodule detection in CT scans. Diagnostics (Basel) 10:131. https://doi.org/10.3390/diagnostics10030131

  26. Weston AD, Korfiatis P, Kline TL et al (2018) Automated abdominal segmentation of CT scans for body composition analysis using deep learning. Radiology 290:669–679. https://doi.org/10.1148/radiol.2018181432

    Article  PubMed  Google Scholar 

  27. Hashimoto F, Kakimoto A, Ota N et al (2019) Automated segmentation of 2D low-dose CT images of the psoas-major muscle using deep convolutional neural networks. Radiol Phys Technol 12:210–215. https://doi.org/10.1007/s12194-019-00512-y

    Article  PubMed  Google Scholar 

  28. Lee H, Troschel FM, Tajmir S et al (2017) Pixel-level deep segmentation: artificial intelligence quantifies muscle on computed tomography for body morphometric analysis. J Digit Imaging 30:487–498. https://doi.org/10.1007/s10278-017-9988-z

    Article  PubMed  PubMed Central  Google Scholar 

  29. Park HJ, Shin Y, Park J et al (2020) Development and validation of a deep learning system for segmentation of abdominal muscle and fat on computed tomography. Korean J Radiol 21:88–100. https://doi.org/10.3348/kjr.2019.0470

    Article  PubMed  Google Scholar 

  30. Chen L-C, Papandreou G, Kokkinos I, et al (2017) DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs. ArXiv160600915 Cs

  31. Belharbi S, Chatelain C, Hérault R et al (2017) Spotting L3 slice in CT scans using deep convolutional network and transfer learning. Comput Biol Med 87:95–103. https://doi.org/10.1016/j.compbiomed.2017.05.018

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

The authors wish to thank Charles-André Cuenod (CA), Gilles Soulat (GS), Mehdi Bouaboula (MB), and Jonas Deidier (JD) from the Radiology Department of the Hôpital européen Georges Pompidou, Assistance Publique-Hôpitaux de Paris, Paris, France for their help in evaluating for the presence of artifacts precluding the use of the segmentation algorithm. Mathilde Parent (MP), Marie Bruneel (MB), Victoria Tortochot (VT) students from the Université de Paris, AP-HP, Hôpital européen Georges Pompidou, Department of Radiology, PARCC UMRS 970, INSERM, Paris, France.

Funding

This study has received funding in part by the French Government under the management of the Agence Nationale de la Recherche as part of the “Investissements d’avenir” program, reference ANR-19-P3IA-0001 (PRAIRIE 3IA Institute).

Author information

Authors and Affiliations

  1. Department of Radiology, Hôpital Européen Georges Pompidou, AP-HP, Université de Paris, PARCC UMRS 970, INSERM, 20 Rue Leblanc, 75015, Paris, France

    Victoire Roblot, Sarah Mezghani, Loïc Duron & Laure Fournier

  2. Foodvisor, Paris, France

    Yann Giret

  3. Department of Medical Oncology, Hôpital Européen Georges Pompidou, AP-HP, Université de Paris, INSERM CIC1418-EC Clinical Epidemiology Team, Paris, France

    Edouard Auclin & Stéphane Oudard

  4. Informatics and Clinical Research Unit, Department of Biostatistics, Hôpital européen Georges Pompidou, AP-HP, Université de Paris, INSERM CIC1418-EC Clinical Epidemiology Team, Paris, France

    Armelle Arnoux

  5. Department of Radiology, Fondation Ophtalmologique Adolphe de Rothschild, Paris, France

    Loïc Duron

Authors
  1. Victoire Roblot
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yann Giret
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Sarah Mezghani
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Edouard Auclin
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Armelle Arnoux
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Stéphane Oudard
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Loïc Duron
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Laure Fournier
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Victoire Roblot.

Ethics declarations

Guarantor

The scientific guarantor of this publication is Laure Fournier, MD PhD.

Conflict of Interest

The authors of this manuscript declare no relationships with any companies whose products or services may be related to the subject matter of the article.

Statistics and Biometry

Armelle Arnoux, PhD, has significant statistical expertise.

She works in the university of Paris, AP-HP, Hôpital européen Georges Pompidou, Department of Biostatistics, Informatics and Clinical Research Unit, INSERM CIC1418-EC Clinical Epidemiology team, Paris, France.

Informed Consent

Written informed consent was waived by the Institutional Review Board.

Ethical Approval

Institutional Review Board approval was obtained, which waived the need for written informed consent. Indeed, in our institutions, the patient gives a global written consent to reuse his/her data for research and education, on condition that an IRB reviews the protocol and its ethical compliance to French law.

The IRB approval has been submitted with the manuscript.

Study subjects or cohorts overlap

Previously to this study, we had developed our diagnostic algorithm in a training population from a data challenge, and you will find enclosed the pdf of the paper referencing the population (Lassau et al, Diagn Interv Imaging 2020), which was one among three data challenges. One of our validation population has been previously published in part, and you will also find the pdf (Auclin et al, Genitourin Cancer 2017). In this previous study, the authors used a semi-automated method to calculate skeletal muscle index, and its purpose was to determine whether sarcopenia was predictive of prognosis and toxicity for metastatic renal cell cancer patients under everolimus. We included 44 of the 124 patients from the initial study, because we needed to have access to the original CT images of patients.

Methodology

• retrospective

• observational

• performed at one institution

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Roblot, V., Giret, Y., Mezghani, S. et al. Validation of a deep learning segmentation algorithm to quantify the skeletal muscle index and sarcopenia in metastatic renal carcinoma. Eur Radiol 32, 4728–4737 (2022). https://doi.org/10.1007/s00330-022-08579-9

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00330-022-08579-9

Keywords

Search

Navigation