Musculoskeletal (MSK) conditions are the leading global contributor to chronic pain and disability, primarily characterized by biomechanical impairment (Blyth et al., 2019, GBD 2016 DALYs and HALE Collaborators, 2017, Piedrahita, 2006). Biomechanical analysis is crucial for identifying abnormal motions that may be caused by or contribute to MSK pathology. Beyond a standard physical examination, clinical tools for measuring biomechanical function include 1) static goniometry (Gajdosik and Bohannon, 1987, Laupattarakasem et al., 1990, van Trijffel et al., 2010), 2) functional testing (de Melo et al., 2022, Kahraman et al., 2016; M. R. Pourahmadi et al., 2018), and 3D skeletal tracking (Lam et al., 2023, do Rosário, 2014). Although useful historically and in some specific situations, static goniometry and functional testing generally fail to capture the complexity of full-body motion and require specialized training, which can affect reliability (van Trijffel et al., 2010) and lack consistency (Finley et al., 2015, Norkin and White, 2016, Reissner et al., 2019). In contrast, contemporary 3D biomechanical analysis leads to a large dataset consisting of three-dimensional data for every landmark across time, highlighting the necessity to extract meaningful and clinically useful metrics.

Traditionally, 3D skeletal tracking data is analyzed by either (1) extracting isolated kinematic/kinetic measures, often using biomechanical models, or (2) reducing the dimensionality of the data to create a single metric, often through Principal Component Analysis (PCA). Isolated kinematic measures are frequently selected from the dataset, but determining whether these measures effectively capture a patient’s overall biomechanical impairment is challenging. For example, in chronic low back pain (LBP) patients, the highest prevalent MSK condition globally (Wu et al., 2020), lumbar range of motion and higher order trunk kinematics (maximum velocity, acceleration) are frequently studied (Laird et al., 2014, Lehman, 2004, Poitras et al., 2000). However, these measures exhibit high variability, conflicting results, neglect time-series data, and overlook meaningful full-body (including lower extremity) compensatory strategies (McGregor et al., 1997, Papi et al., 2018, Song et al., 2012). Their usefulness may be limited to specific patients, activities, or time points, whereas growing evidence supports a full-body approach for a more comprehensive assessment of biomechanical function (Papi et al., 2018).

On the other hand, kinematic scores aimed to condense 3D biomechanical data into a single metric have been developed for gait (Herrera-Valenzuela et al., 2022, Massaad et al., 2014, Schwartz and Rozumalski, 2008), balance (Chang et al., 2020, Eveleigh et al., 2023, Halvorson et al., 2022), and upper extremity motion (Jurkojć et al., 2017). Further, many scores highlight deviations from healthy controls, suggesting promising clinical utility. Although these scores provide a valuable foundation for our work, they also present several limitations, including one or more of the following: lack of compatibility with other activities, insufficient coverage of full-body posture, lack of dynamic analysis by neglecting time series data, and dependence on patient sample size. Due to the volume of motion capture data, PCA is widely used for dimensionality reduction and calculating these kinematic scores. PCA transforms original data into a set of uncorrelated components that capture the most significant variance, highlighting meaningful patterns within datasets. PCA is typically employed across patients over kinematic measurements at select or single time points (Brandon et al., 2013, Deluzio and Astephen, 2007, Halvorson et al., 2022, Keller et al., 2022, Warmenhoven et al., 2021). Although this approach has shown to be useful in a research context, it challenges clinical utility as applying PCA across patients conflicts patient-specific movements with global population movement patterns that might not reflect any individual subject. An additional challenge with across-patient PCA and other machine-learning techniques is the requirement for large sample sizes to compute a stable metric. To address these limitations, PCA can be applied at the individual level, as previously demonstrated in other fields and some kinematic analyses [28]. However, existing research in kinematics primarily focuses on describing motion patterns rather than developing a composite metric for clinical utility.

Therefore, in line with a more personalized assessment of MSK impairments, the objective of this study was to detail the methodology of the K-Score algorithm and evaluate its strengths and limitations. We aimed to explore its potential applications as a quantitative approach for assessing differences in postural movement patterns, defined by alignment with healthy control motion, across multiple body landmarks and time. Posture reflects the spatial relationship of the body, is a primary mode of compensation, and is fundamental to understanding differences in biomechanical function and motor control (Park et al., 2023; M. Pourahmadi et al., 2023, Sung and Lee, 2024). Further, the K-Score addresses the challenges of current approaches by capturing full-body motion that can be used across diverse activities, incorporating dynamic time-series data, and providing a comprehensive, single metric that does not depend on patient sample size, ensuring potentially broad applicability for both clinical and research contexts. We conducted a comparative analysis to evaluate the advantages and limitations of the K-Score algorithm in relation to traditional isolated kinematic metrics and across-patient PCA analysis. All three approaches were applied to identify differences within patients with chronic LBP, which was selected due to its high prevalence (Wu et al., 2020) and tremendous heterogeneity in movement patterns (Alsubaie et al., 2023).

We hypothesized that the K-Score metric would effectively capture the coordinated interactions across multiple body segments over time that quantify differences in overall movement patterns, while maintaining independence from sample size and demonstrating robustness against demographic variability. This would make it more effective at distinguishing patient types and reducing the limitations typically associated with traditional isolated metrics and the conventional PCA method. In both clinical and research settings, the K-Score could be applied to evaluate movement impairments across different patient populations, monitor treatment effectiveness and recovery, and identify movement adaptations that contribute to long-term MSK health.