The face blurring of images plays a key role in protecting privacy. However, in computer vision, especially for the human pose estimation task, machine-learning models are currently trained, ...validated, and tested on original datasets without face blurring. Additionally, the accuracy of human pose estimation is of great importance for kinematic analysis. This analysis is relevant in areas such as occupational safety and clinical gait analysis where privacy is crucial. Therefore, in this study, we explore the impact of face blurring on human pose estimation and the subsequent kinematic analysis. Firstly, we blurred the subjects’ heads in the image dataset. Then we trained our neural networks using the face-blurred and the original unblurred dataset. Subsequently, the performances of the different models, in terms of landmark localization and joint angles, were estimated on blurred and unblurred testing data. Finally, we examined the statistical significance of the effect of face blurring on the kinematic analysis along with the strength of the effect. Our results reveal that the strength of the effect of face blurring was low and within acceptable limits (<1°). We have thus shown that for human pose estimation, face blurring guarantees subject privacy while not degrading the prediction performance of a deep learning model.
Purpose
This in vitro biomechanical study compares residual lumbar range of motion (ROM) and rod strain after lumbopelvic instrumentation using 2 rods, 4 rods and interbody cages.
Methods
Seven human ...cadaveric specimens were instrumented from L1 to sacrum, and pelvic screws were implanted. The pelvis was constrained and moments up to 7.5 Nm were applied to T12. Segmental L1–S1 ROM was analyzed by tracking radiopaque balls implanted in each vertebra using biplanar radiographs. Deformation within principal rods was measured by strain gauges. Four configurations were compared: 2 rods (2R), 4 rods (4R), 4 rods + ALIF at L4–L5 and L5–S1 (4R + ALIF), 2 rods + ALIF (2R + ALIF).
Results
Intact average global L1–S1 ROM was 42.9° (27.9°–66.0°) in flexion–extension (FE), 35.2° (26.8°–51.8°) in lateral bending (LB), 18.6° (6.7°–47.8°) in axial rotation (AR). In FE, average ROM was 1.9° with both 4-rod configurations versus 2.5° with 2R and 2.8° with 2R + ALIF (
p
< 0.05). In LB, ROM ranged between 1.2° and 1.5° without significant differences. In AR, ROM was 2.5° with both 4-rod configurations versus 2.9° with 2R (
p
= 0.07) and 3.1° with 2R ALIF (
p
= 0.01). In FE, strain decreased by 64% and 65% in principal rods at L3–L4 with 4-rod. When comparing 2-rod configurations, strain decreased by 1% in flexion and increased by 22% in extension at L3–L4 when adding an ALIF at L4–L5 and L5–S1.
Conclusions
Double rods and interbody cages decrease residual ROM in FE and AT. Double rods seem efficient in limiting strain in principal rods. The use of single rods with cages at the lumbosacral junction increases strain at the first adjacent level without cage.
Shear wave elastography (SWE) is an ultrasound technique to obtain soft tissue mechanical properties. The aim of this study was to establish the reliability of SWE in young children, define reference ...data on healthy ones and compare the shear modulus of healthy and spastic muscles from cerebral palsy (CP). The reproducibility is evaluated: at rest, on 7 children without any musculoskeletal pathology by 3 different operators, on 2 muscles: biceps brachii long head and medial gastrocnemius. The comparison study was made, on the same 2 muscles, at rest and under passive stretching, with a control group (29 healthy children), a spastic group (spastic muscles of 16 children from CP) and a non-spastic group (non-spastic muscles of 14 children from CP). The intra-operator reliability and inter-operator reliability, in terms of standard deviation, were 0.6 kPa (11.2% coefficient of variation (CV)) and 0.8 kPa (14.9% CV) for the biceps, respectively, and 0.4 kPa (11.5% CV) and 0.5 kPa (13.8% CV) for the gastrocnemius. At rest, no significant difference was found. Under passive stretching, the non-spastic CP biceps were significantly stiffer than the control ones (p = 0.033). Spastic gastrocnemius had a higher shear modulus than in the control muscles (p = 0.0003) or the non-spastic CP muscles (p = 0.017). CP stretched medial gastrocnemius presented an abnormally high shear moduli for 50% of patients.
Development of shear wave elastography gave access to non-invasive muscle stiffness assessment in vivo. The aim of the present study was to define a measurement protocol to be used in clinical ...routine for quantifying the shear modulus of lower limb muscles. Four positions were defined to evaluate shear modulus in 10 healthy subjects: parallel to the fibers, in the anterior and posterior aspects of the lower limb, at rest and during passive stretching. Reliability was first evaluated on two muscles by three operators; these measurements were repeated six times. Then, measurement reliability was compared in 11 muscles by two operators; these measurements were repeated three times. Reproducibility of shear modulus was 0.48 kPa and repeatability was 0.41 kPa, with all muscles pooled. Position did not significantly influence reliability. Shear wave elastography appeared to be an appropriate and reliable tool to evaluate the shear modulus of lower limb muscles with the proposed protocol.
BackgroundThere is an increasing recognition of the clinical importance of the sagittal plane alignment of the spine. A prospective study of several radiographic parameters of the sagittal profile of ...the spine was conducted to determine the physiological values of these parameters, to calculate the variations of these parameters according to epidemiological and morphological data, and to study the relationships among all of these parameters.MethodsSagittal radiographs of the head, spine, and pelvis of 300 asymptomatic volunteers, made with the subject standing, were evaluated. The following parameters were measuredlumbar lordosis, thoracic kyphosis, T9 sagittal offset, sacral slope, pelvic incidence, pelvic tilt, intervertebral angulation, and vertebral wedging angle from T9 to S1. The radiographs were digitized, and all measurements were performed with use of a software program. Two different analyses, a descriptive analysis characterizing these parameters and a multivariate analysis, were performed in order to study the relationships among all of them.ResultsThe mean values (and standard deviations) were 60° ± 10° for maximum lumbar lordosis, 41° ± 8.4° for sacral slope, 13° ± 6° for pelvic tilt, 55° ± 10.6° for pelvic incidence, and 10.3° ± 3.1° for T9 sagittal offset. A strong correlation was found between the sacral slope and the pelvic incidence (r = 0.8); between maximum lumbar lordosis and sacral slope (r = 0.86); between pelvic incidence and pelvic tilt (r = 0.66); between maximum lumbar lordosis and pelvic incidence, pelvic tilt, and maximum thoracic kyphosis (r = 0.9); and, finally, between pelvic incidence and T9 sagittal offset, sacral slope, pelvic tilt, maximum lumbar lordosis, and thoracic kyphosis (r = 0.98). The T9 sagittal offset, reflecting the sagittal balance of the spine, was dependent on three separate factorsa linear combination of the pelvic incidence, maximum lumbar lordosis, and sacral slope; the pelvic tilt; and the thoracic kyphosis.Conclusions and Clinical RelevanceThis description of the physiological spinal sagittal balance should serve as a baseline in the evaluation of pathological conditions associated with abnormal angular parameter values. Before a patient with spinal sagittal imbalance is treated, the reciprocal balance between various spinal angular parameters needs to be taken into account. The correlations between angular parameters may also be useful in calculating the corrections to be obtained during treatment.
Purpose
To explore 3D hip orientation in standing position in subjects with adult spinal deformity (ASD) presenting with different levels of compensatory mechanisms.
Methods
Subjects with ASD (
n
= ...159) and controls (
n
= 68) underwent full-body biplanar X-rays with the calculation of 3D spinopelvic, postural and hip parameters. ASD subjects were grouped as ASD with knee flexion (ASD-KF) if they compensated by flexing their knees (knee flexion ≥ 5°), and ASD with knee extension (ASD-KE) otherwise (knee flexion < 5°). Spinopelvic, postural and hip parameters were compared between the three groups. Univariate and multivariate analyses were then computed between spinopelvic and hip parameters.
Results
ASD-KF had higher SVA (67 ± 66 mm vs. 2 ± 33 mm and 11 ± 21 mm), PT (27 ± 14° vs. 18 ± 9° and 11 ± 7°) and PI-LL mismatch (20 ± 26° vs − 1 ± 18° and − 13 ± 10°) when compared to ASD-KE and controls (all
p
< 0.05). ASD-KF also had a more tilted (34 ± 11° vs. 28 ± 9° and 26 ± 7°), anteverted (24 ± 6° vs. 20 ± 5° and 18 ± 4°) and abducted (59 ± 6° vs. 57 ± 4° and 56 ± 4°) acetabulum, with a higher posterior coverage (100 ± 6° vs. 97 ± 7° for ASD-KE) when compared to ASD-KE and controls (all
p
< 0.05). The main determinants of acetabular tilt, acetabular abduction and anterior acetabular coverage were PT, SVA and LL (adjusted R
2
0.12; 0.5).
Conclusions
ASD subjects compensating with knee flexion have altered hip orientation, characterized by increased posterior coverage (acetabular anteversion, tilt and posterior coverage) and decreased anterior coverage which can together lead to posterior femoro-acetabular impingement, thus limiting pelvic retroversion. This underlying mechanism could be potentially involved in the hip-spine syndrome.
Highlights • The repeatability and uncertainty of GDI (Gillette Deviation Index) were evaluated. • Repeatability coefficient obtained on GDI for typically developing children was ±10. • GDI is ...robust, not sensitive to the noise applied on its entries. • We found a moderate correlation between GDI and GMFCS (Gross Motor Function Classification System).
Prospective study of 131 patients and volunteers recruited for an analysis of spinal alignment and gravity line (GL) assessment by force plate analysis.
To determine relationships between GL, foot ...position, and spinopelvic landmarks in subjects with varying sagittal alignment. Additionally, the study sought to analyze the role of the pelvis in the maintenance of GL position.
Force plate technology permits analysis of foot position and GL in relation to radiographically obtained landmarks. Previous investigation noted fixed GL-heel relationship across a wide age range despite changes in thoracic kyphosis. The pelvis as balance regulator has not been studied in the setting of sagittal spinal deformity.
The 131 subjects were grouped by sagittal vertical axis (SVA) offset from the sacrum: sagittal forward (>2.5 cm), neutral (-2.5 cm <or= SVA <or= 2.5 cm), and sagittal backward (SVA <-2.5 cm). Simultaneous spinopelvic radiographs and GL measure were obtained. Offsets between spinopelvic landmarks, heel position, and GL were calculated. Group comparisons were made for all offsets to determine significance.
Aside from the offset T9-GL and GL-heels, all other offsets between spinopelvic landmarks and GL revealed significant differences (P < 0.001) across the 3 subject groups. However, with increasing SVA, the GL kept a rather fixed location relative to the feet. A correlation between posterior pelvic shift in relation to the heels with increasing SVA in this study population was confirmed (r = 0.6, P < 0.001).
Increasing SVA in standing subjects leads to a posterior pelvic shift in relation to the feet. However, no significant difference in GL-heel offset is noted with increasing SVA. It thus appears that pelvic shift (in relation to the feet) is an important component in maintaining a rather fixed GL-Heels offset even in the setting of variable SVA and trunk inclination.
Purpose
The aim of this work is to propose a classification algorithm to automatically detect treatment for scoliosis (brace, implant or no treatment) in postero-anterior radiographs. Such automatic ...labelling of radiographs could represent a step towards global automatic radiological analysis.
Methods
Seven hundred and ninety-six frontal radiographies of adolescents were collected (84 patients wearing a brace, 325 with a spinal implant and 387 reference images with no treatment). The dataset was augmented to a total of 2096 images. A classification model was built, composed by a forward convolutional neural network (CNN) followed by a discriminant analysis; the output was a probability for a given image to contain a brace, a spinal implant or none. The model was validated with a stratified tenfold cross-validation procedure. Performance was estimated by calculating the average accuracy.
Results
98.3% of the radiographs were correctly classified as either reference, brace or implant, excluding 2.0% unclassified images. 99.7% of brace radiographs were correctly detected, while most of the errors occurred in the reference group (i.e. 2.1% of reference images were wrongly classified).
Conclusion
The proposed classification model, the originality of which is the coupling of a CNN with discriminant analysis, can be used to automatically label radiographs for the presence of scoliosis treatment. This information is usually missing from DICOM metadata, so such method could facilitate the use of large databases. Furthermore, the same model architecture could potentially be applied for other radiograph classifications, such as sex and presence of scoliotic deformity.
PDF
