By: 4 October 2022
EUROSPINE 2022: Best in show

Get acquainted with four of the abstract submissions for EUROSPINE 2022 in Milan, which have been elected as Best of Show and will be presented in the Best of Show and Award papers session


Lumbar intervertebral discs from childhood to adulthood: a 26-year longitudinal MRI study with special reference to low back pain 

Presentation by Teija Lund, Martina Lohman, Anni Aavikko, Leena Ristolainen, Hannu Kautiainen, Dietrich Schlenzka 

Background: The reported life-time occurrence of low back pain (LBP) in children and adolescents varies from 7% to 72% reaching adult levels by the end of puberty. Although most children and adolescents report no LBP, 16 to 37% suffer occasional bouts, and up to 10 % report repeated episodes. Childhood LBP has been proposed a significant risk factor for LBP in adulthood.  

 Disc degeneration (DD) on MRI becomes more prevalent with age and is frequently seen in asymptomatic individuals.  A significant increase in DD has been reported after the pubertal growth spurt with up to 44 % of adolescents with LBP demonstrating signs of DD on MRI. The value of MRI changes in predicting future LBP seems low. However, DD on MRI in early adulthood has been shown to predispose the individual for a more rapid progression of degeneration.  

Purpose: In the present study, our objective was 1) to describe the natural history of lumbar intervertebral discs from childhood to adulthood, and 2) to investigate whether findings of DD are associated with LBP. Hereby we report on the association of signs of DD on MRI after the pubertal growth spurt to self-reported LBP at the age of 18 and 34. 

Materials and Methods: In 1994, we recruited 94 healthy 8-year-old school children for a semi-structured interview, a clinical examination, and a lumbar spine MRI with follow-ups at the ages of 11 (n=81) and 18 (n=71). In January 2021, at the age of 34, all those subjects whose contact details were known (n=89) were invited for a long-term follow-up. 48 participants went through a semi-structured interview, a clinical examination and a 1.5T lumbar spine MRI. From the MRI investigations we assessed the signal intensity of the lumbar discs visually using the Pfirrmann classification and calculated the Pfirrmann Summary Score (PSS) for the whole lumbar spine by adding up the scores of the individual discs for a range of 5-25. Further, we analyzed the association of PSS at the age of 18 to self-reported LBP at the ages of 18 and 34.  

Results: At the age of 8, 11 and 18, disc changes on MRI occurred in 18, 10, and 38% of the participants, respectively. By the age of 34, 73% of the participants had developed at least one disc with a Pfirrmann grade 3 or higher. At the age of 18, the occurrence of self-reported LBP was 50% reaching 73% by the age of 34. The mean PSS significantly increased from the age of 18 to the age of 34 from 10.4 (SD 1.6) to 11.6 (SD 1.7) with a mean change of 1.2 (95% CI 0.7 to 1.7). No statistically significant difference in PSS was noticed at the ages of 18 or 34 between participants with or without LBP. The OR (95% CI) of PSS at the age of 18 with reference to LBP at the age of 18 was 1.32 (0.91 to 1.92), and to LBP at the age of 34 1.49 (0.93 to 2.38). 

Conclusion: More widespread lumbar DD at the age of 18 marginally predicted LBP in adulthood with every 1-point increase in PSS increasing the risk of LBB 1.5-fold. Thus, accelerated disc degeneration after the pubertal growth spurt may predispose the individual for LBP in adulthood. Our results might offer some cue towards distinguishing “normal” age-related versus “pathologic” DD. 


Curve progression and Health Related Quality of Life (HRQoL) in idiopathic scoliosis – 40-year follow-up from diagnosis 

Presentation by Casper Dragsted, Lærke Ragborg, Søren Ohrt-Nissen, Thomas Andersen, Martin Gehrchen, Benny Dahl, Spine Unit, Department of Orthopaedic Surgery, Rigshospitalet and University of Copenhagen 

Introduction: Treatment of idiopathic scoliosis in childhood is mainly guided by curve size and location with the aim to prevent curve progression and long-term effects of larger deformities. It is generally accepted that curves >50° will progress throughout adulthood, but less well described what happens with mild to moderate curves and how curve size and progression is related to Health Related Quality of Life (HRQoL) at long term follow-up. Moreover, there are conflicting results regarding risk of progression in thoracic and thoracolumbar/lumbar (TL/L) main curves. A few older but well-recognized studies have described the natural history of idiopathic scoliosis leading to recommendations regarding spinal fusion, but it is less well described what happens with patients with mild to moderate scoliosis who do not undergo spinal fusion. 

The question is whether some of these patients experience curve progression to at point where surgery in adolescence would have been preferable? And moreover, if curve progression and main curve size reflects HRQoL in adulthood? 

We have had the unique opportunity to assess curve progression and HRQoL in a cohort of patients 40 years after a diagnosis of idiopathic scoliosis in childhood. All patients were seen at the Department of Orthopaedic surgery at Rigshospitalet in Copenhagen during the years 1972-1983 and diagnosed with scoliosis. We were able to retrieve the old patient files including detailed descriptions of treatment, curve size and location. Patients were either observed, treated with a Boston Brace or subsequently underwent spondylodesis with Harrington rod instrumentation. 

All patients were invited for a follow-up visit including a clinical examination, long standing full-spine radiographs and answered Euro-Qol 5D-3L and SRS-22r questionnaires. The patients also underwent extended pulmonary function tests and clinical examination. These results will be published in future studies. A total of 104 of all eligible patients completed follow-up (69%), 91 of these were diagnosed with juvenile or adolescent idiopathic scoliosis and assessed in this study. 

The purpose of the study was to assess long-term curve progression and HRQoL in patients with idiopathic scoliosis and compare TL/L and thoracic curves. 


The Artificial Intelligence Motion study (AIM): Artificial Intelligence-assisted image recognition of cervical spine motion

Presentation by Valérie Schuermans, Sara El-Ateif, Soroosh Poorgholi, Paul Algra, Henk van Santbrink, Toon Boselie 

Analysing motion in the spine remains a challenge. Spinal motion is commonly analysed through segmental range of motion (sROM). However, this method shows high intra- and inter-individual variability. Instead of investigating quantitative motion of the cervical spine (sROM) we focus on qualitative motion analysis. 

In a previous study by our group, motion patterns in radiographic recordings were investigated. The sequence of segmental contributions is determined by calculating the relative rotations of each segment. Areas of interest of the occiput (C0) and cervical vertebrae (C1-C7) were manually drawn, after which normalised gradient fields were used to track the vertebrae throughout the recordings. In the following step, more precise contours were drawn and manually corrected in each frame of the recording. This technique has been proven to be sensitive and specific for the analysis of motion patterns, but unfortunately is extremely labor-intensive.   

The AIM-study uses the annotated data develop artificial intelligence (AI) – assisted segmentation models to recognise and track cervical vertebrae.  

Radiographic flexion-extension recordings of healthy volunteers and pre-operative patients with degenerative disc disease were used. The contours of C0 to C7 were manually annotated in all frames of the recordings. 

To achieve the localisation of vertebrae across multiple frames, two segmentation approaches were developed using ResUnet++ network and Mask-RCNN. Finally, parts of both methods were combined in an optimal algorithm. The final model was trained on 2025 images and validated in 176 images for C0-C7.  

The AI-model almost exactly identifies cervical vertebral contours in images that have not been ‘seen’ by the algorithm before. Intersection over union (IOU) was measured to compare the ground truth with the AI-predicted contour, and proved of high accuracy in all recordings. Consistency of a specific contour throughout the individual recordings was measured through the mean IOU of consecutive frames, for the ground truth and AI-prediction separately. The consistency of contours throughout recordings was highly comparable between the human annotation and AI prediction.  

The developed AI-model enables time-efficient and accurate segmentation of C0-C7. Moreover, human experience and training is not required. The accessibility of this method allows extensive analysis of spinal motion, of which in depth knowledge is currently still lacking. We are currently developing a method to calculate motion patterns based on the segmented cervical vertebrae. The next steps will be to validate the model with recordings of patients with anatomical variants or implants in-situ. This can provide new insights in the quality of cervical spine motion in individuals. 

Ultimately, the aim is to investigate the relationship between motion of the cervical spine and the development of pathology.  


Lumbar segmental mobility and degenerative phenotypes: a large study based on an automated image analysis tool 

Presentation by Fabio Galbusera, Francine Mariaux, Stephan Bernhart, Sebastian Winklhofer, Hans-Joachim Wilke, François Porchet, Anne F. Mannion 

Synopsis: It is well known that lumbar disc degeneration impacts spine mobility, but the importance of the specific degenerative phenotypes on lumbar motion has never been quantified so far. Our AI-based tool allowed us to automatically analyse a large dataset of flexion-extension radiographs to determine how the single degenerative findings as observed in MRI affect the segmental range of motion. 

Text: The mobility and stability of the lumbar spine have been considered one of the key factors in the degenerative cascade for decades, since they have major implications in terms of the biomechanics of the pathological segments and, in turn, on pain and disability. For this reason, the lumbar mobility has been thoroughly investigated in a large number of in vitro and experimental studies as well as in vivo studies by performing measurements on radiographic images. Nevertheless, the manual measurement of the segmental mobility in flexion-extension radiographs suffers from inaccuracies and repeatability issues that can reach magnitudes comparable to those of the quantities under investigation. Computer-aided tools, allowing for improved repeatability and accuracy, are currently considered the state-of-the-art but also have limited reproducibility, and are time-consuming as well. We recently developed and validated a computer program able to automatically calculate the motion parameters and used it to investigate the association between overall disc degeneration and spinal motion. In this work we explored the potential correlations between the lumbar mobility and specific lumbar degeneration phenotypes, namely disc degeneration, endplate defects, Modic changes and spondylolisthesis, as observed on MRI. To this aim, we exploited a large database of flexion-extension radiographs of 592 patients suffering from lumbar spondylolisthesis who also underwent MRI examination. The presence of endplate defects, Modic changes, spondylolisthesis as well as the Pfirrmann grade of degeneration were annotated by an expert radiologist at all lumbar levels. All flexion-extension radiographs were processed with the automated tool to calculate the segmental motion parameters. Differences in spinal motion between degenerative phenotype grades were examined using standard statistical methods. In general, the segmental motion generally decreased with progressive increases in disc degeneration grade. The range of motion was significantly smaller in patients with endplate defects, Modic changes and spondylolisthesis, whereas lower anteroposterior translation during flexion-extension was seen only with Modic changes. In summary, this study confirmed previous findings of an inverse relationship between the extent of disc degeneration and spinal segmental mobility and provided information about the relevance of specific degenerative phenotypes which was not available so far. The study provided data which are very valuable for companies involved in the development of motion-preserving implants as well as for the basic research community, such as for example for the validation of the numerical and experimental models that are commonly used for the investigation of spine pathologies.  


For more information about EUROSPINE 2022 and to book your place, visit: