Enhancing biomechanical machine learning with limited data: generating realistic synthetic posture data using generative artificial intelligence.

Objective: Biomechanical Machine Learning (ML) models, particularly deep-learning models, demonstrate the best performance when trained using extensive datasets. However, biomechanical data are frequently limited due to diverse challenges. Effective methods for augmenting data in developing ML models, specifically in the human posture domain, are scarce. Therefore, this study explored the feasibility of leveraging generative artificial intelligence (AI) to produce realistic synthetic posture data by utilizing three-dimensional posture data. Methods: Data were collected from 338 subjects through surface topography. A Variational Autoencoder (VAE) architecture was employed to generate and evaluate synthetic posture data, examining its distinguishability from real data by domain experts, ML classifiers, and Statistical Parametric Mapping (SPM). The benefits of incorporating augmented posture data into the learning process were exemplified by a deep autoencoder (AE) for automated feature representation. Results: Our findings highlight the challenge of differentiating synthetic data from real data for both experts and ML classifiers, underscoring the quality of synthetic data. This observation was also confirmed by SPM. By integrating synthetic data into AE training, the reconstruction error can be reduced compared to using only real data samples. Moreover, this study demonstrates the potential for reduced latent dimensions, while maintaining a reconstruction accuracy comparable to AEs trained exclusively on real data samples. Conclusion: This study emphasizes the prospects of harnessing generative AI to enhance ML tasks in the biomechanics domain.

Influence of Lateral Sitting Wedges on the Rasterstereographically Measured Scoliosis Angle in Patients Aged 10-18 Years with Adolescent Idiopathic Scoliosis.

Adolescent idiopathic scoliosis (AIS) is a three-dimensional axial deviation of the spine diagnosed in adolescence. Despite a long daily sitting duration, there are no studies on whether scoliosis can be positively influenced by sitting on a seat wedge. For the prospective study, 99 patients with AIS were measured with the DIERS formetric III 4D average, in a standing position, on a level seat and with three differently inclined seat wedges (3°, 6° and 9°). The rasterstereographic parameters 'scoliosis angle' and 'lateral deviation RMS' were analysed. The side (ipsilateral/contralateral) on which the optimal correcting wedge was located in relation to the lumbar/thoraco-lumbar convexity was investigated. It was found that the greatest possible correction of scoliosis occurred with a clustering in wedges with an elevation on the ipsilateral side of the convexity. This clustering was significantly different from a uniform distribution (p < 0.001; chi-square = 35.697 (scoliosis angle); chi-square = 54.727 (lateral deviation RMS)). It should be taken into account that the effect of lateral seat wedges differs for individual types of scoliosis and degrees of severity. The possibility of having a positive effect on scoliosis while sitting holds great potential, which is worth investigating in follow-up studies.

Creation and Evaluation of a Severity Classification of Hyperkyphosis and Hypolordosis for Exercise Therapy.

The rise in the occurrence of musculoskeletal disorders, such as thoracic hyperkyphosis (THK) or lumbar hypolordosis (LHL), is a result of demographic changes. Exercise therapy is an effective approach that can reduce related disabilities and costs. To ensure successful therapy, an individualized exercise program adapted to the severity of the disorder is expedient. Nevertheless, appropriate classification systems are scarce. This project aimed to develop and evaluate a severity classification focused on exercise therapy for patients with THK or LHL. A multilevel severity classification was developed and evaluated by means of an online survey. Reference limits of spinal shape angles were established by data from video rasterstereography of 201 healthy participants. A mean kyphosis angle of 50.03° and an average lordosis angle of 40.72° were calculated as healthy references. The strength of the multilevel classification consisting of the combination of subjective pain and objective spinal shape factors was confirmed by the survey (70% agreement). In particular, the included pain parameters were considered relevant by 78% of the experts. Even though the results of the survey provide important evidence for further analyses and optimization options of the classification system, the current version is still acceptable as therapeutic support.

Reference values and functional descriptions of transverse plane spinal dynamics during gait based on surface topography.

Spinal dynamics during gait have been of interest in research for many decades. Based on respective previous investigations, the pelvis is generally expected to be maximally forward rotated on the side of the reference leg at the beginning of each gait cycle and to reach its maximum counterrotation approximately at the end of the reference leg's stance phase. The pelvic-upper-thoracic-spine coordination converges towards an anti-phase movement pattern in high velocities during ambulation. The vertebral bodies around the seventh thoracic vertebra are considered to be an area of transition during human ambulation where no or at least little rotary motion can be observed. The respective cranial and caudal vertebrae meanwhile are expected to rotate conversely around this spinal point of intersection. However, these previous assumptions are based on scarce existing research, whereby only isolated vertebrae have been analyzed contemporaneously. Due to huge methodological differences in data capturing approaches, the results are additionally hardly comparable to each other and involved measurement procedures are often not implementable in clinical routines. Furthermore, none of the above-mentioned methods provided reference data for spinal motion during gait based on an appropriate number of healthy participants. Hence, the aim of this study was to present such reference data for spinal rotary motion of every vertebral body from C7 down to L4 and the pelvis derived from surface topographic back shape analyses in a cohort of 201 healthy participants walking on a treadmill at a given walking speed of 5 km/h. Additionally, the spine's functional movement behavior during gait should be described in the transverse plane based on data derived from this noninvasive, clinically suitable measurement approach and, in conclusion, the results shall be compared against those of previous research findings derived from other measurement techniques. Contrary to the previous functional understanding, the area of the mid-thoracic spine was found to demonstrate the largest amplitude of rotary motion of all investigated vertebrae and revealed an approximately counterrotated movement behavior compared to the rotary motion of the pelvis. In both directions, spinal rotation during gait seemed to be initiated by the pelvis. The overlying vertebrae followed in succession in the sense of an ongoing movement. Therefore, the point of intersection was not statically located in a specific anatomical section of the spine. Instead, it was found to be dynamic, ascending from one vertebra to the next from caudal to cranial in dependence of the pelvis's rotation initiation.

Reference Values for 3D Spinal Posture Based on Videorasterstereographic Analyses of Healthy Adults.

Visual examinations are commonly used to analyze spinal posture. Even though they are simple and fast, their interrater reliability is poor. Suitable alternatives should be objective, non-invasive, valid and reliable. Videorasterstereography (VRS) is a corresponding method that is increasingly becoming established. However, there is a lack of reference data based on adequate numbers of participants and structured subgroup analyses according to sex and age. We used VRS to capture the spinal posture of 201 healthy participants (aged 18-70 years) divided into three age cohorts. Three-dimensional reference data are presented for the global spine parameters and for every vertebral body individually (C7-L4) (here called the specific spine parameters). The vertebral column was found to be systematically asymmetric in the transverse and the coronal planes. Graphical presentations of the vertebral body posture revealed systematic differences between the subgroups; however, large standard deviations meant that these differences were not significant. In contrast, several global parameters (e.g., thoracic kyphosis and lumbar lordosis) indicated differences between the analyzed subgroups. The findings confirm the importance of presenting reference data not only according to sex but also according to age in order to map physiological posture changes over the life span. The question also arises as to whether therapeutic approximations to an almost symmetrical spine are biomechanically desirable.

Machine learning techniques demonstrating individual movement patterns of the vertebral column: the fingerprint of spinal motion.

Surface topography systems enable the capture of spinal dynamic movement; however, it is unclear whether vertebral dynamics are unique enough to identify individuals. Therefore, in this study, we investigated whether the identification of individuals is possible based on dynamic spinal data. Three different data representations were compared (automated extracted features using contrastive loss and triplet loss functions, as well as simple descriptive statistics). High accuracies indicated the possible existence of a personal spinal 'fingerprint', therefore enabling subject recognition. The present work forms the basis for an objective comparison of subjects and the transfer of the method to clinical use cases.

Evaluation of 3D vertebral and pelvic position by surface topography in asymptomatic females: presentation of normative reference data.

BACKGROUND: Deviations from a conventional physiologic posture are often a cause of complaint. According to current literature, the upright physiological spine posture exhibits inclinations in the sagittal plane but not in the coronal and transverse planes, but individual vertebral body positions of asymptomatic adults have rarely been described using surface topography. Therefore, this work aims to form a normative reference dataset for the thoracic and lumbar vertebral bodies and for the pelvis in all three planes in asymptomatic women. METHODS: In a prospective, cross-sectional, monocentric study, 100 pain-free asymptomatic women, aged 20-64 years were enrolled. Habitual standing positions of the trunk were measured using surface topography. Data were analyzed in all three planes. Age sub-analysis was: 1) ages ≤ 40 years and 2) ages ≥ 41 years. Two-sample t-tests were used for age comparisons of the vertebral bodies, vertebra prominence (VP)-L4, and global parameters. One-sample t-tests were used to test deviations from symmetrical zero positions of VP-L4. RESULTS: Coronal plane: on average, the vertebral bodies were tilted to the right between the VP and T4 (maximum: T2 - 1.8° ± 3.2), while between T6 and T11 they were tilted to the left (maximum: T7 1.1° ± 1.9). T5 and L2 were in a neutral position, overall depicting a mean right-sided lateral flexion from T2 to T7 (apex at T5). Sagittal plane: the kyphotic apex resided at T8 with - 0.5° ± 3.6 and the lumbar lordotic apex at L3 with - 2.1° ± 7.4. Transverse plane: participants had a mean vertebral body rotation to the right ranging from T6 to L4 (maximum: T11 - 2.2° ± 3.5). Age-specific differences were seen in the sagittal plane and had little effect on overall posture. CONCLUSIONS: Asymptomatic female volunteers standing in a habitual posture displayed an average vertebral rotation and lateral flexion to the right in vertebral segments T2-T7. The physiological asymmetrical posture of women could be considered in spinal therapies. With regard to spinal surgery, it should be clarified whether an approximation to an absolutely symmetrical posture is desirable from a biomechanical point of view? This data set can also be used as a reference in clinical practice. TRIAL REGISTRATION: This study was registered with WHO (INT: DRKS00010834) and approved by the responsible ethics committee at the Rhineland-Palatinate Medical Association (837.194.16).

Classification and Automated Interpretation of Spinal Posture Data Using a Pathology-Independent Classifier and Explainable Artificial Intelligence (XAI).

Clinical classification models are mostly pathology-dependent and, thus, are only able to detect pathologies they have been trained for. Research is needed regarding pathology-independent classifiers and their interpretation. Hence, our aim is to develop a pathology-independent classifier that provides prediction probabilities and explanations of the classification decisions. Spinal posture data of healthy subjects and various pathologies (back pain, spinal fusion, osteoarthritis), as well as synthetic data, were used for modeling. A one-class support vector machine was used as a pathology-independent classifier. The outputs were transformed into a probability distribution according to Platt's method. Interpretation was performed using the explainable artificial intelligence tool Local Interpretable Model-Agnostic Explanations. The results were compared with those obtained by commonly used binary classification approaches. The best classification results were obtained for subjects with a spinal fusion. Subjects with back pain were especially challenging to distinguish from the healthy reference group. The proposed method proved useful for the interpretation of the predictions. No clear inferiority of the proposed approach compared to commonly used binary classifiers was demonstrated. The application of dynamic spinal data seems important for future works. The proposed approach could be useful to provide an objective orientation and to individually adapt and monitor therapy measures pre- and post-operatively.

General method for automated feature extraction and selection and its application for gender classification and biomechanical knowledge discovery of sex differences in spinal posture during stance and gait.

Modern technologies enable to capture multiple biomechanical parameters often resulting in relational data. The current work proposes a generally applicable method comprising automated feature extraction, ensemble feature selection and classification to best capture the potentials of the data also for generating new biomechanical knowledge. Its benefits are demonstrated in the concrete biomechanically and medically relevant use case of gender classification based on spinal data for stance and gait. Very good results for accuracy were obtained using gait data. Dynamic movements of the lumbar spine in sagittal and frontal plane and of the pelvis in frontal plane best map gender differences.