Early prediction of curve progression to surgical threshold in Adolescent Idiopathic Scoliosis with bone microarchitecture phenotyping – a step closer to Precision Medicine
Invited conference paper presented and published in conference proceedings
CUHK Authors
Full Text
There are no full text file(s) associated with this record. |
Other information
AbstractBackground:
Adolescent Idiopathic Scoliosis is a complex 3-dimensional deformity of the spine. Skeletal maturity, curve magnitude and curve type are commonly used to guide clinical decision for observation, bracing or surgery. Active ongoing research for an effective prognostication model that can accurately predict curve progression, thus allowing timely management for groups with high risk of curve progression, while avoid unnecessary overtreatment in those with low risk is an important common goal.
In multifactorial etiopathogenic AIS patients, bone quality has shown potential to be a significant prognostic factor in curve progression. High-resolution-peripheral-quantitative-computed-tomography (HRpQCT) is now available for low radiation, 3-dimensional skeletal evaluation and can generate 11 quantitative bone quality parameters. However, to perform composite analysis on these complex parameters can be difficult and time-consuming. Previous studies were limited by investigating only bone mineral density without combined evaluation of different bone qualities, there is a need to search for a novel way to perform composite analysis of HRpQCT-generated bone parameters and explore their associations with curve progression in AIS patients.
Our study therefore aims to (a) utilise unsupervised machine-learning technology to analyse and identify hidden Bone phenotype clusters amongst HRpQCT-generated quantitative parameters, (b) investigate the Bone phenotype’s association with curve progression and progression to surgical threshold in a 6-year longitudinal cohort.
Methods:
101 AIS girls (age 12.26+/-0.87) were recruited and followed up longitudinally till skeletal maturity (age 18.47 +/- 0.96). Anthropometric data, curve magnitude, and bone quality parameters by HRpQCT were documented at their first clinic visit. Their skeletal maturity was staged by Thumb Ossification Composite Index (TOCI) and were all at peripubertal peak height velocity. Average baseline major curve cobb angle was 25.95+/- 9deg, and their HRpQCT-bone quality parameters were analysed and clustered by Fuzzy-CMeans - an unsupervised machine learning model.
Results:
In our cohort, three bone microarchitecture phenotype clusters were identified by our machine learning model. Patients with Phenotype-1 had normal bone characteristics. Phenotype-2 was characterised by significantly lower periosteal perimeter and trabecular area, Phenotype-3 had significantly lower cortical bone quality parameters, trabecular bone volume fraction, and trabecular-thickness. During the 6 years longitudinal follow up, their respective bone-microarchitecture phenotype was found to be persistent till skeletal maturity, and Phenotype-3 was associated with a significantly increased risk of curve progression to >50deg, which is our surgical threshold (Odds Ratio (OR)= 4.88, p= 0.029).
Conclusion:
With AI machine learning algorithm, three bone microarchitecture phenotypes can be identified in peripubertal AIS girls at first clinic visit. Those with bone microarchitecture phenotype-3 has a significant risk of curve progression to surgical threshold (>50deg) compared to phenotype-1 and 2. Bone phenotype-2 and 3 were associated with significantly increased risk of curve progression compared to normal bone phenotype.
The three phenotypes in our cohort have similar baseline skeletal maturity and scoliotic curve severity, we found the key difference to curve progression was their bone quality and microarchitecture. Further ongoing expanded multicentre validation of our prognostication model and clinical trials could have significant clinical implication in the management and prevention of curve progression for early AIS patients.
Adolescent Idiopathic Scoliosis is a complex 3-dimensional deformity of the spine. Skeletal maturity, curve magnitude and curve type are commonly used to guide clinical decision for observation, bracing or surgery. Active ongoing research for an effective prognostication model that can accurately predict curve progression, thus allowing timely management for groups with high risk of curve progression, while avoid unnecessary overtreatment in those with low risk is an important common goal.
In multifactorial etiopathogenic AIS patients, bone quality has shown potential to be a significant prognostic factor in curve progression. High-resolution-peripheral-quantitative-computed-tomography (HRpQCT) is now available for low radiation, 3-dimensional skeletal evaluation and can generate 11 quantitative bone quality parameters. However, to perform composite analysis on these complex parameters can be difficult and time-consuming. Previous studies were limited by investigating only bone mineral density without combined evaluation of different bone qualities, there is a need to search for a novel way to perform composite analysis of HRpQCT-generated bone parameters and explore their associations with curve progression in AIS patients.
Our study therefore aims to (a) utilise unsupervised machine-learning technology to analyse and identify hidden Bone phenotype clusters amongst HRpQCT-generated quantitative parameters, (b) investigate the Bone phenotype’s association with curve progression and progression to surgical threshold in a 6-year longitudinal cohort.
Methods:
101 AIS girls (age 12.26+/-0.87) were recruited and followed up longitudinally till skeletal maturity (age 18.47 +/- 0.96). Anthropometric data, curve magnitude, and bone quality parameters by HRpQCT were documented at their first clinic visit. Their skeletal maturity was staged by Thumb Ossification Composite Index (TOCI) and were all at peripubertal peak height velocity. Average baseline major curve cobb angle was 25.95+/- 9deg, and their HRpQCT-bone quality parameters were analysed and clustered by Fuzzy-CMeans - an unsupervised machine learning model.
Results:
In our cohort, three bone microarchitecture phenotype clusters were identified by our machine learning model. Patients with Phenotype-1 had normal bone characteristics. Phenotype-2 was characterised by significantly lower periosteal perimeter and trabecular area, Phenotype-3 had significantly lower cortical bone quality parameters, trabecular bone volume fraction, and trabecular-thickness. During the 6 years longitudinal follow up, their respective bone-microarchitecture phenotype was found to be persistent till skeletal maturity, and Phenotype-3 was associated with a significantly increased risk of curve progression to >50deg, which is our surgical threshold (Odds Ratio (OR)= 4.88, p= 0.029).
Conclusion:
With AI machine learning algorithm, three bone microarchitecture phenotypes can be identified in peripubertal AIS girls at first clinic visit. Those with bone microarchitecture phenotype-3 has a significant risk of curve progression to surgical threshold (>50deg) compared to phenotype-1 and 2. Bone phenotype-2 and 3 were associated with significantly increased risk of curve progression compared to normal bone phenotype.
The three phenotypes in our cohort have similar baseline skeletal maturity and scoliotic curve severity, we found the key difference to curve progression was their bone quality and microarchitecture. Further ongoing expanded multicentre validation of our prognostication model and clinical trials could have significant clinical implication in the management and prevention of curve progression for early AIS patients.
All Author(s) ListLau A, Yang G, Hung A, Lee W, Arnubrat K, Wan R, Lam T, Cheng J
Name of ConferenceAsia Pacific Spine Society (APSS) 7th Annual Meeting 2024
Start Date of Conference13/06/2024
End Date of Conference16/06/2024
Place of ConferenceHong Kong
Country/Region of ConferenceHong Kong
Year2024
LanguagesEnglish-United Kingdom
