A new functional classification in adult spinal deformity patients based on 3D gait analysis

Patients with adult spinal deformity (ASD) are known to have a deteriorated quality of life that is usually evaluated through disability questionnaires [1]. While classical evaluation of ASD is based on static standing X-rays, 3D gait analysis (3DGA) was shown to be important in assessing patient’s functional disabilities [2,3]. However, its interpretation can be a burden on spine surgeons due to the large number of kinematic parameters that are calculated. A simplified classification based on few kinematic parameters would be beneficial to evaluate functional disabilities in the setting of spinal deformity. To develop a new functional classification for ASD patients based on 3DGA. 157 subjects (110 ASD and 47 controls, age and sex matched: 54±18 years, 73% F) filled the Oswestry Disability Index (ODI) questionnaire (score:0 to 100, increases with disability). All subjects underwent 3DGA from which the difference between kinematic parameters and normative values were calculated for the: head (ODHA angle between center of the head, center of hips and the vertical), thorax, spine segments, pelvis and lower limbs as well as time-distance parameters (normalized step length, walking speed) [4,5]. In order to investigate the main kinematic parameters that determine gait variability between patients, a principal component analysis (PCA) was computed, followed by a correlation analysis with ODI scores. Then, in order to develop a functional classification based on 3DGA, a decision tree classifier was applied (training/testing data: 80%/20%) to determine cutoffs of main kinematics that predict classes of ODI: class-1 <20 (mild), class-2 [20–40] (intermediate), class-3 >40 (severe). ROC curves, accuracy and F1-scores were calculated to evaluate the model. The PCA explained 50% of data variance with the main kinematic parameters being step length, ODHA, and pelvic tilt. ODHA and step length were correlated to ODI (r=0.40, -0.70 resp.; p<0.05). ASD patients were classified as follows based on their ODI: N= 36, 42, 32 in classes 1, 2 and 3 respectively. The model predicted ODI classes with the following cutoffs: class-1 for ODHA<3° and step length>0.34; class-2 for ODHA [3-4°[ and step length [0.32-0.34[; class-3 for pelvic tilt≤3° or ODHA≥4° and step length<0.32. The accuracy of the model was 82%, with an AUC of 0.9. The F1-score of the classifier was 0.89 with the following scores for classes 1, 2 and 3: 0.87, 0.89 and 0.93 respectively (Fig. 1). This study showed that the head position (ODHA), pelvic tilt and step length allow to delineate and predict disability levels, making this functional evaluation clinically relevant. The 3 main kinematic parameters are sufficient to be representative of patients’ kinematics and their functional disability. This new functional classification can help surgeons in evaluating ASD patients in clinical routine and will be used to evaluate surgical intervention on patient’s functional changes. Fig. 1: Functional classification in adult spinal deformity using kinematic parameters and its evaluation based on ROC curve.Download : Download high-res image (59KB)Download : Download full-size image