Correlation of texture analysis of paraspinal musculature on MRI with different clinical endpoints: Lumbar Stenosis Outcome Study (LSOS)
The aim of this study was to apply texture analysis (TA) on paraspinal musculature in T2-weighted (T2w) magnetic resonance images (MRI) of symptomatic lumbar spinal stenosis (LSS) patients and correlate the findings with clinical outcome measures.
Ninety patients were prospectively enrolled in the multi-centric Lumbar Stenosis Outcome Study (LSOS). All patients received a T2w MRI, from which we selected axial images perpendicular to the intervertebral disc at level L3/4 for TA. Regions-of-interest (ROI) were drawn of the paraspinal musculature and 304 TA features/ ROI were calculated. As clinical outcome measurements, we analysed three commonly applied measures: Spinal Stenosis Measure (SSM), Roland-Morris Disability Questionnaire (RMDQ), as well as the Numeric Rating Scale (NRS). We used two machine learning-based classifiers: Decision table, and k-nearest neighbours (k-NN).
We observed no meaningful correlation between TA in paraspinal musculature and the two clinical outcome measures SSM symptoms and SSM function, while a moderate correlation was observed regarding the outcome measures RMDQ (k-NN: r = 0.56) and NRS (Decision Table: r = 0.72).
In conclusion, MR TA is a viable tool to quantify medical images and illustrate correlations of microarchitectural changes invisible to a human reader with potential clinical impact.
• TA is feasible on paraspinal musculature using MRI.
• TA on paraspinal musculature correlates with SSM and RMDQ.
• TA may enable a statement regarding clinical impact of imaging findings.
KeywordsMagnetic resonance imaging Machine learning Spine Muscles
Area under the curve
Cross-sectional muscle area
Digital imaging and communications in medicine
Grey-level co-occurrence matrix
Lumbar back pain
Lumbar spinal stenosis
Magnetic resonance imaging
Numeric rating scale (pain)
Picture archiving and communication system
Region of interest
Spinal stenosis measure
Turbo spin echo
This study has received funding by Helmut Horten Foundation, Baugarten Foundation, Pfizer-Foundation for geriatrics & research in geriatrics, Symphasis Charitable Foundation and OPO Foundation funds.
Compliance with ethical standards
The scientific guarantor of this publication is Roman Guggenberger, MD.
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
Ulrike Held and Jakob M. Burgstaller kindly provided statistical advice for this manuscript.
Written informed consent was obtained from all subjects (patients) in this study.
Institutional Review Board approval was obtained.
Study subjects or cohorts overlap
• multicentre study
- 4.Fortin M, Lazary A, Varga PP, Battie MC (2017) Association between paraspinal muscle morphology, clinical symptoms and functional status in patients with lumbar spinal stenosis. Eur Spine J. https://doi.org/10.1007/s00586-017-5228-y
- 6.de Bruin F, ter Horst S, Bloem HL et al (2016) Prevalence of degenerative changes of the spine on magnetic resonance images and radiographs in patients aged 16-45 years with chronic back pain of short duration in the Spondyloarthritis Caught Early (SPACE) cohort. Rheumatology (Oxford) 55:56–65CrossRefGoogle Scholar
- 14.Keller A, Gunderson R, Reikeras O, Brox JI (2003) Reliability of computed tomography measurements of paraspinal muscle cross-sectional area and density in patients with chronic low back pain. Spine (Phila Pa 1976) 28:1455–1460Google Scholar
- 15.Ingrisch M, Schneider MJ, Norenberg D et al (2017) Radiomic analysis reveals prognostic information in T1-weighted baseline magnetic resonance imaging in patients with glioblastoma. Investig Radiol. https://doi.org/10.1097/RLI.0000000000000349
- 17.Kolossvary M, Kellermayer M, Merkely B, Maurovich-Horvat P (2017) Cardiac computed tomography radiomics: a comprehensive review on radiomic techniques. J Thorac Imaging. https://doi.org/10.1097/RTI.0000000000000268
- 20.De Certaines J, Larcher T, Duda D et al (2015) Application of texture analysis to muscle MRI: 1-What kind of information should be expected from texture analysis? EPJ Nonlinear Biomedical Physics 3Google Scholar
- 22.Prasanna P, Patel J, Partovi S, Madabhushi A, Tiwari P (2016) Radiomic features from the peritumoral brain parenchyma on treatment-naive multi-parametric MR imaging predict long versus short-term survival in glioblastoma multiforme: preliminary findings. Eur Radiol. https://doi.org/10.1007/s00330-016-4637-3
- 26.Lisson CS, Lisson CG, Flosdorf K et al (2017) Diagnostic value of MRI-based 3D texture analysis for tissue characterisation and discrimination of low-grade chondrosarcoma from enchondroma: a pilot study. Eur Radiol. https://doi.org/10.1007/s00330-017-5014-6
- 27.Bresnahan LE, Smith JS, Ogden AT et al (2016) Assessment of paraspinal muscle cross-sectional area following lumbar decompression: minimally invasive versus open approaches. Clin Spine Surg. https://doi.org/10.1097/BSD.0000000000000038
- 28.Ulrich NH, Burgstaller JM, Held U et al (2017) The influence of single-level versus multilevel decompression on the outcome in multisegmental lumbar spinal stenosis: analysis of the Lumbar Spinal Outcome Study (LSOS) data. Clin Spine Surg. https://doi.org/10.1097/BSD.0000000000000469
- 40.Crawford RJ, Cornwall J, Abbott R, Elliott JM (2017) Manually defining regions of interest when quantifying paravertebral muscles fatty infiltration from axial magnetic resonance imaging: a proposed method for the lumbar spine with anatomical cross-reference. BMC Musculoskelet Disord 18:25CrossRefGoogle Scholar
- 41.Baessler B, Mannil M, Oebel S, Maintz D, Alkadhi H, Manka R (2017) Subacute and chronic left ventricular myocardial scar: accuracy of texture analysis on nonenhanced cine MR images. Radiology. https://doi.org/10.1148/radiol.2017170213:170213
- 42.Dobbin KK, Simon RM (2011) Optimally splitting cases for training and testing high dimensional classifiers. BMC Med Genet 4:31Google Scholar
- 43.Dwyer B, Hutchings K (1977) Flowchart optimisation in cope, a multi-choice decision table. Aust Comp J 9:92Google Scholar
- 44.Altman NS (1992) An introduction to kernel and nearest-neighbor nonparametric regression. Am Stat 46:175–185Google Scholar
- 45.Mannil M, Burgstaller JM, Thanabalasingam A et al (2018) Texture analysis of paraspinal musculature in MRI of the lumbar spine: analysis of the Lumbar Stenosis Outcome Study (LSOS) data. Skeletal Radiol. https://doi.org/10.1007/s00256-018-2919-3
- 48.Leinonen V, Maatta S, Taimela S et al (2003) Paraspinal muscle denervation, paradoxically good lumbar endurance, and an abnormal flexion-extension cycle in lumbar spinal stenosis. Spine (Phila Pa 1976) 28:324–331Google Scholar