Search

Abstract Deep learning-based image segmentation has allowed for the fully automated, accurate, and rapid analysis of musculoskeletal (MSK) structures from medical images. However, current approaches were either applied only to 2D cross-sectional images, addressed few structures, or were validated on small datasets, which limit the application in large-scale databases. This study aimed to validate an improved deep learning model for volumetric MSK segmentation of the hip and thigh with uncertainty estimation from clinical computed tomography (CT) images. Databases of CT images from multiple manufacturers/scanners, disease status, and patient positioning were used. The segmentation accuracy, and accuracy in estimating the structures volume and density, i.e., mean HU, were evaluated. An approach for segmentation failure detection based on predictive uncertainty was also investigated. The model has improved all segmentation accuracy and structure volume/density evaluation metrics compared to a shallower baseline model with a smaller training database (N = 20). The predictive uncertainty yielded large areas under the receiver operating characteristic (AUROC) curves (AUROCs ≥ .95) in detecting inaccurate and failed segmentations. Furthermore, the study has shown an impact of the disease severity status on the model’s predictive uncertainties when applied to a large-scale database. The high segmentation and muscle volume/density estimation accuracy and the high accuracy in failure detection based on the predictive uncertainty exhibited the model’s reliability for analyzing individual MSK structures in large-scale CT databases.

Abstract The effect of gravity on the lungs has been evaluated using computed tomography (CT) in the supine and prone positions but not the standing position. However, as humans spend most of the daytime in the standing position, we aimed to compare lung attenuation gradients between the supine and standing positions, and to assess the correlations between the lung attenuation gradients and participant characteristics, including pulmonary function test results. Overall, 100 healthy participants underwent conventional/supine and upright CT, and lung attenuation gradients were measured. Lung attenuation gradients in anteroposterior direction were greater in the supine position than in standing position (all p values

Abstract This study aimed to reproduce and analyse the in vivo dynamic rotational motion of the forearm and to clarify forearm motion involvement and the anatomical function of the interosseous membrane (IOM). The dynamic forearm rotational motion of the radius and ulna was analysed in vivo using a novel image-matching method based on fluoroscopic and computed tomography images for intensity-based biplane two-dimensional–three-dimensional registration. Twenty upper limbs from 10 healthy volunteers were included in this study. The mean range of forearm rotation was 150 ± 26° for dominant hands and 151 ± 18° for non-dominant hands, with no significant difference observed between the two. The radius was most proximal to the maximum pronation relative to the ulna, moved distally toward 60% of the rotation range from maximum pronation, and again proximally toward supination. The mean axial translation of the radius relative to the ulna during forearm rotation was 1.8 ± 0.8 and 1.8 ± 0.9 mm for dominant and non-dominant hands, respectively. The lengths of the IOM components, excluding the central band (CB), changed rotation. The transverse CB length was maximal at approximately 50% of the rotation range from maximum pronation. Summarily, this study describes a detailed method for evaluating in vivo dynamic forearm motion and provides valuable insights into forearm kinematics and IOM function.

Aims: This study aimed to develop and validate a fully automated system that quantifies proximal femoral bone mineral density (BMD) from CT images. Methods: The study analyzed 978 pairs of hip CT and dual-energy X-ray absorptiometry (DXA) measurements of the proximal femur (DXA-BMD) collected from three institutions. From the CT images, the femur and a calibration phantom were automatically segmented using previously trained deep-learning models. The Hounsfield units of each voxel were converted into density (mg/cm3). Then, a deep-learning model trained by manual landmark selection of 315 cases was developed to select the landmarks at the proximal femur to rotate the CT volume to the neutral position. Finally, the CT volume of the femur was projected onto the coronal plane, and the areal BMD of the proximal femur (CT-aBMD) was quantified. CT-aBMD correlated to DXA-BMD, and a receiver operating characteristic (ROC) analysis quantified the accuracy in diagnosing osteoporosis. Results: CT-aBMD was successfully measured in 976/978 hips (99.8%). A significant correlation was found between CT-aBMD and DXA-BMD (r = 0.941; p < 0.001). In the ROC analysis, the area under the curve to diagnose osteoporosis was 0.976. The diagnostic sensitivity and specificity were 88.9% and 96%, respectively, with the cutoff set at 0.625 g/cm2. Conclusion: Accurate DXA-BMD measurements and diagnosis of osteoporosis were performed from CT images using the system developed herein. As the models are open-source, clinicians can use the proposed system to screen osteoporosis and determine the surgical strategy for hip surgery. Cite this article: Bone Joint Res 2023;12(9):589–596.

Abstract This paper presents methods of decomposition of musculoskeletal structures from radiographs into multiple individual muscle and bone structures. While existing solutions require dual-energy scan for the training dataset and are mainly applied to structures with high-intensity contrast, such as bones, we focused on multiple superimposed muscles with subtle contrast in addition to bones. The decomposition problem is formulated as an image translation problem between (1) a real X-ray image and (2) multiple digitally reconstructed radiographs, each of which contains a single muscle or bone structure, and solved using unpaired training based on the CycleGAN framework. The training dataset was created via automatic computed tomography (CT) segmentation of muscle/bone regions and virtually projecting them with geometric parameters similar to the real X-ray images. Two additional features were incorporated into the CycleGAN framework to achieve a high-resolution and accurate decomposition: hierarchical learning and reconstruction loss with the gradient correlation similarity metric. Furthermore, we introduced a new diagnostic metric for muscle asymmetry directly measured from a plain X-ray image to validate the proposed method. Our simulation and real-image experiments using real X-ray and CT images of 475 patients with hip diseases suggested that each additional feature significantly enhanced the decomposition accuracy. The experiments also evaluated the accuracy of muscle volume ratio measurement, which suggested a potential application to muscle asymmetry assessment from an X-ray image for diagnostic and therapeutic assistance. The improved CycleGAN framework can be applied for investigating the decomposition of musculoskeletal structures from single radiographs.

Abstract This study presents a novel framework for classifying and visualizing pneumonia induced by COVID-19 from CT images. Although many image classification methods using deep learning have been proposed, in the case of medical image fields, standard classification methods are unable to be used in some cases because the medical images that belong to the same category vary depending on the progression of the symptoms and the size of the inflamed area. In addition, it is essential that the models used be transparent and explainable, allowing health care providers to trust the models and avoid mistakes. In this study, we propose a classification method using contrastive learning and an attention mechanism. Contrastive learning is able to close the distance for images of the same category and generate a better feature space for classification. An attention mechanism is able to emphasize an important area in the image and visualize the location related to classification. Through experiments conducted on two-types of classification using a three-fold cross validation, we confirmed that the classification accuracy was significantly improved; in addition, a detailed visual explanation was achieved comparison with conventional methods.

Abstract The purpose of the study was to develop a deep learning network for estimating and constructing highly accurate 3D bone models directly from actual X-ray images and to verify its accuracy. The data used were 173 computed tomography (CT) images and 105 actual X-ray images of a healthy wrist joint. To compensate for the small size of the dataset, digitally reconstructed radiography (DRR) images generated from CT were used as training data instead of actual X-ray images. The DRR-like images were generated from actual X-ray images in the test and adapted to the network, and high-accuracy estimation of a 3D bone model from a small data set was possible. The 3D shape of the radius and ulna were estimated from actual X-ray images with accuracies of 1.05 ± 0.36 and 1.45 ± 0.41 mm, respectively.

Protein localization in cells has been analyzed by fluorescent labeling using indirect immunofluorescence and fluorescent protein tagging. However, the relationships between the localization of different proteins had not been analyzed using artificial intelligence. Here, we applied convolutional networks for the prediction of localization of the cytoskeletal proteins from the localization of the other proteins. Lamellipodia are one of the actin-dependent subcellular structures involved in cell migration and are mainly generated by the Wiskott-Aldrich syndrome protein (WASP)-family verprolin homologous protein 2 (WAVE2) and the membrane remodeling I-BAR domain protein IRSp53. Focal adhesion is another actin-based structure that contains vinculin protein and promotes lamellipodia formation and cell migration. In contrast, microtubules are not directly related to actin filaments. The convolutional network was trained using images of actin filaments paired with WAVE2, IRSp53, vinculin, and microtubules. The generated images of WAVE2, IRSp53, and vinculin were highly similar to their real images. In contrast, the microtubule images generated from actin filament images were inferior without the generation of filamentous structures, suggesting that microscopic images of actin filaments provide more information about actin-related protein localization. Collectively, this study suggests that image translation by the convolutional network can predict the localization of functionally related proteins, and the convolutional network might be used to describe the relationships between the proteins by their localization.

Purpose: Pelvic position on the sagittal plane is usually evaluated with the pelvic sagittal inclination (PSI) angle from a single radiograph. However, the reproducibility of pelvic positioning has not been investigated, and thus, the validity of measuring the PSI from a single film/time point is not understood. Herein, the reproducibility of a patient’s pelvic positions in supine and standing postures was analyzed. Methods: A total of 34 patients who underwent either a pelvic osteotomy or total hip arthroplasty were enrolled in this study. Preoperative radiographs in both supine and standing postures were acquired twice (first X-ray and second X-ray) within 6 months; preoperative computed tomography (CT) images of the full pelvis were also acquired in a supine posture (preop-CT). To eliminate measurement variability, each PSI was automatically measured from radiographs and CT images through the use of CT segmentation and landmark localization followed by intensity-based 2D-3D registration. The absolute difference of PSI among each image was calculated and the intra-class correlation coefficient (ICC) in each posture was also analyzed. Results: The median absolute differences of PSI in the supine posture were 1.3° between the first and second X-rays, 1.2° between the first X-ray and preop-CT, and 1.3° between the second X-ray and preop-CT. The median absolute difference of PSI in the standing posture was 1.5°. The ICC was 0.965 (95% CI: 0.939–0.981) in supine and 0.977 (95% CI: 0.954–0.988) during standing. Conclusions: Pelvic positions in supine and standing postures are reproducible. Thus, measuring the PSI from a single radiograph is reliable.

Purpose: Intertrochanteric fractures are usually treated with open reduction and internal fixation, but controversy still remains regarding the proper placement of the lag screw on the anteroposterior view. The stability of the lag screw has been shown to correlate with the bone quality around the screw, but the three-dimensional distribution of the bone mineral density (BMD) in the femoral head has not been studied in detail. Herein, the BMD along the femoral neck axis was measured to clarify the recommended position of the lag screw. Methods: Ten femoral heads acquired from intertrochanteric fractures were evaluated in this study. Each femoral head was scanned with micro computed tomography and the BMD along the femoral neck axis was measured in five regions: center, anterior, posterior, superior, and inferior. The BMD on the anteroposterior view (superior, center, and inferior) and the BMD on the lateral view (anterior, center, and posterior) were compared. Results: The BMD of the center region (173.0 ± 50.6 mg/cm 3 ) was significantly higher than that of the inferior region (139.7 ± 50.1 mg/cm 3 ) on the anteroposterior view ( p < 0.01). On the lateral view, the BMD was lower than the center region in the anterior region (165.7 ± 52.8 mg/cm 3 ) and in the posterior region (157.5 ± 42.3 mg/cm 3 ), but the difference was not significant. Conclusion: The BMD was higher in the center region of the femoral head than in the inferior region. Therefore, lag screws are recommended to be inserted into the center of the femoral head.

Femoroplasty is a potential preventive treatment for osteoporotic hip fractures. It involves augmenting mechanical properties of the femur by injecting Polymethylmethacrylate (PMMA) bone cement. To reduce the risks involved and maximize the outcome, however, the procedure needs to be carefully planned and executed. An important part of the planning system is predicting infiltration of cement into the porous medium of cancellous bone. We used the method of Smoothed Particle Hydrodynamics (SPH) to model the flow of PMMA inside porous media. We modified the standard formulation of SPH to incorporate the extreme viscosities associated with bone cement. Darcy creeping flow of fluids through isotropic porous media was simulated and the results were compared with those reported in the literature. Further validation involved injecting PMMA cement inside porous foam blocks - osteoporotic cancellous bone surrogates - and simulating the injections using our proposed SPH model. Millimeter accuracy was obtained in comparing the simulated and actual cement shapes. Also, strong correlations were found between the simulated and the experimental data of spreading distance (R(2) = 0.86) and normalized pressure (R(2) = 0.90). Results suggest that the proposed model is suitable for use in an osteoporotic femoral augmentation planning framework.