Spinal sagittal alignment: investigation of postoperative pelvic kinematic improvement in patients with spinal sagittal imbalance using machine learning methods

Madadi, Sadegh; Rostami, Mostafa; Farahani, Hadi; Nikouee, Farshad; Samadian, Mohammad; Haddas, Ram

doi:10.22060/ajmc.2025.23866.1317

Spinal sagittal alignment: investigation of postoperative pelvic kinematic improvement in patients with spinal sagittal imbalance using machine learning methods

Articles in Press

Document Type : Special Issue: MLKD 2024

Authors

Sadegh Madadi ¹

Mostafa Rostami ¹

Hadi Farahani ²

Farshad Nikouee ³

Mohammad Samadian ⁴

Ram Haddas ⁵

¹ Department of Biomedical Engineering, Amirkabir University of Technology, Tehran, Iran

² Department of Computer and Data Sciences, Faculty of Mathematical Sciences, Shahid Beheshti University, Tehran, Iran

³ Bone and Joint Reconstruction Research Center, Shafa Orthopedic Hospital, Iran University of Medical Sciences, Tehran, Iran

⁴ Department of Neurosurgery, Shahid Beheshti University of Medical Sciences, Tehran, Iran

⁵ University of Rochester Medical Center, Rochester, NY, USA

10.22060/ajmc.2025.23866.1317

Abstract

Background: The pelvic plays an important role in human movement, and it is the foundation that provides stability during activities such as walking. Abnormal condition of the pelvic area, whether it is an abnormality of alignment or function, requires timely treatment intervention. Traditionally, pelvic examination has been performed through static two-dimensional imaging with very limited insight into real-time pelvic dynamics. Inertial measurement unit (IMU) sensors are already very powerful in acquiring all nuances of movement mechanics; with the addition of ML techniques, they can serve as an effective methodology for diagnosing pelvic movement patterns for different activities. Material and Methods: This study investigates the gait pattern of 50 female patients with spinal sagittal imbalance (SSI) compared to 50 controls. Various machine learning (ML) models were applied using IMU data collected during gait analysis in order to identify and assess abnormalities in movement. Results: Results: While the Support Vector Machine (SVM) achieved the highest classification accuracy (99.07%) in identifying pelvic movement disorders using IMU data, the Linear Discriminant Analysis (LDA) model showed the most balanced performance with the highest F1-score (99.00%) and precision (98.00%). This indicates that LDA may be preferable in settings where minimizing false positives and achieving balanced classification is critical.

Keywords

Machine learning

Spine

Gait analysis

Classification

Subjects