3D Imaging Research Articles

Enhanced 3D imaging based on regional optical texture synthesis

Optical information synthesis, which fuses LiDAR and optical cameras, has the potential for highly detailed 3D representations. However, due to the disparity of information density between point clouds and images, conventional matching methods based on points often lose significant information. To address this issue, we propose a regional matching method to bridge the differences in information density between point clouds and images. In detail, fine semantic regions are extracted from images by analyzing their gradients. Simultaneously, point clouds are transformed into meshes, where each facet corresponds to a coarse semantic region. Extrinsic matrices are used to unify the point cloud coordinate system with the image coordinate system. The mesh is further subdivided based on the guidance of image texture information to create regional matching units. Within each matching unit, the information density of the point cloud and the image is carefully balanced at a semantic level. The texture features of the image are well preserved in the transformed mesh structure. Consequently, the proposed texture synthesis method significantly enhances the overall quality and realism of the 3D imaging.

3D imaging of underwater scanning photon counting Lidar based on multiscale spatio-temporal resolution

3D imaging of underwater scanning photon counting Lidar based on multiscale spatio-temporal resolution

3D fluorescence staining and confocal imaging of low amount of intestinal organoids (enteroids): Protocol accessible to all

The emerging field of 3D organ modeling encounters several imaging issues in particular related to antigen retrieval and sample loss during staining processes. Due to their compact shape, several antibodies fail to penetrate intact organoids or spheroids. Histology of organoids can be approached by paraffin inclusion and sectioning at 5 μm as performed for biopsies. However, to fully understand organoid behavior, including cellular organization, extracellular matrix structure, and their response to treatments, 3D imaging is essential. Here we propose an easy workflow allowing (1) immunostaining with a HIER step, (2) preservation of the intact shape of the organoids, (3) sample immobilization in a focal plane reachable for high resolution/short working distance lenses, and (4) minimizing the risk of loss of precious material.

Deep learning-based polarization 3D imaging method for underwater targets

The significant absorption and scattering of light during its propagation in water severely degrade the quality of underwater imaging, presenting challenges for developing high-precision 3D imaging techniques based on optical methods. Polarization imaging has demonstrated effectiveness in mitigating the effects of scattering, making it a valuable approach for underwater imaging. Additionally, the polarization state of reflected light can be utilized for surface normal estimation and 3D shape reconstruction. This paper presents a learning-based method for 3D shape reconstruction of underwater targets using shape from polarization techniques. To address the lack of publicly available datasets for underwater polarization 3D imaging, we have developed a data acquisition system that simulates Jerlov Type I water conditions, creating a dataset of underwater polarized images along with corresponding ground truth surface normal images. Furthermore, we propose a network framework based on Attention U2Net for the 3D reconstruction of underwater polarized images. This framework is designed to capture detailed texture information of underwater targets and incorporates an effective polarization representation to resolve azimuthal ambiguity, thus enhancing the accuracy of underwater 3D imaging. Experimental results demonstrate that our method effectively addresses azimuthal ambiguity, reduces texture loss during reconstruction, and improves the accuracy of surface normal estimation, achieving superior performance compared to existing methods.

Machine learning in ocular oncology and oculoplasty: Transforming diagnosis and treatment

In the domains of ocular oncology and oculoplasty, machine learning (ML) has become a game-changing technology, providing previously unheard-of levels of precision in diagnosis, treatment planning, and outcome prediction. Using imaging modalities, genomic data, and clinical characteristics, this chapter investigates the integration of machine learning algorithms in the detection and treatment of ocular tumours, including retinoblastoma and uveal melanoma. Through predictive modelling and real-time decision-making, it also emphasises how ML might improve surgical outcomes in oculoplasty, including orbital reconstruction and eyelid correction. Automated examination of fundus photographs, histological slides, and 3D imaging has been made possible by methods like deep learning and natural language processing, which have improved individualised therapeutic approaches and decreased diagnostic errors. Additionally, the use of augmented reality and machine learning in robotics and surgery is a significant development in precision oculoplasty. Notwithstanding its potential, issues including data heterogeneity, algorithm interpretability, and ethical considerations are significant roadblocks that need to be addressed. This chapter explores cutting-edge developments, real-world uses, and potential future paths, offering researchers and doctors a thorough resource.Dipali Vikas Mane, Associate Professor, Shriram Shikshan Sanstha’s College of Pharmacy, Paniv-413113

GMmorph: dynamic spatial matching registration model for 3D medical image based on gated Mamba

Abstract Objective: Deformable registration aims to achieve nonlinear alignment of image space by estimating a dense displacement field. It is commonly used as a preprocessing step in clinical and image analysis applications, such as surgical planning, diagnostic assistance, and surgical navigation. We aim to overcome these challenges: Deep learning-based registration methods often struggle with complex displacements and lack effective interaction between global and local feature information. They also neglect the spatial position matching process, leading to insufficient registration accuracy and reduced robustness when handling abnormal tissues.
Approach: We propose a dual-branch interactive registration model architecture from the perspective of spatial matching. Implicit regularization is achieved through a consistency loss, enabling the network to balance high accuracy with a low folding rate. We introduced the Dynamic Matching Module (DMM) between the two branches of the registration, which generates learnable offsets based on all the tokens across the entire resolution range of the base branch features. Using trilinear interpolation, the model adjusts its feature expression range according to the learned offsets, capturing highly flexible positional differences. To facilitate the spatial matching process, we designed the Gated Mamba Layer (GML) to globally model pixel-level features by associating all voxel information, while the Detail Enhancement Module (DEM), which is based on channel and spatial attention, enhances the richness of local feature details.
Main results: Our study explores the model’s performance in single-modal and multi-modal image registration, including normal brain, brain tumor, and lung images. We propose unsupervised and semi-supervised registration modes and conduct extensive validation experiments. The results demonstrate that the model achieves state-of-the-art performance across multiple datasets..
Significance: By introducing a novel perspective of position matching, the model achieves precise registration of various types of medical data, offering significant clinical value in medical applications.

Posttraining Network Compression for 3D Medical Image Segmentation: Reducing Computational Efforts via Tucker Decomposition.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To investigate whether the computational effort of 3D CT-based multiorgan segmentation with TotalSegmentator can be reduced via Tucker decomposition-based network compression. Materials and Methods In this retrospective study, Tucker decomposition was applied to the convolutional kernels of the TotalSegmentator model, an nnU-Net model trained on a comprehensive CT dataset for automatic segmentation of 117 anatomic structures. The proposed approach reduced the floating-point operations (FLOPs) and memory required during inference, offering an adjustable trade-off between computational efficiency and segmentation quality. This study utilized the publicly available TotalSegmentator dataset containing 1228 segmented CTs and a test subset of 89 CTs, employing various downsampling factors to explore the relationship between model size, inference speed, and segmentation accuracy, evaluated using the Dice score. Results The application of Tucker decomposition to the TotalSegmentator model substantially reduced the model parameters and FLOPs across various compression ratios, with limited loss in segmentation accuracy. Up to 88% of the model's parameters were removed, with no evidence of differences in performance compared with the original model for 113 of 117 classes after fine-tuning. Practical benefits varied across different graphics processing unit architectures, with more distinct speed-ups on less powerful hardware. Conclusion The study demonstrates that posthoc network compression via Tucker decomposition presents a viable strategy for reducing the computational demand of medical image segmentation models without substantially impacting model accuracy. ©RSNA, 2025.

Open Chrono-Morph viewer: visualize big bioimage time series containing heterogeneous volumes

Abstract Summary Time-lapse 3D imaging is fundamental for studying biological processes but requires software able to handle terabytes of voxel data. Although many multidimensional viewing applications exist, they mostly lack support for heterogeneous voxel counts, datatypes, and modalities in a single timeline. Open Chrono-Morph Viewer provides a straightforward graphical user interface to quickly investigate multi-timescale datasets represented as separate volume files in the common NRRD format for compatibility between toolchains. It features dynamic clipping surfaces for rapid investigation of 3D morphology and a scriptable animation API for quantitative, repeatable, publication-quality visualization. It is implemented in pure Python using common libraries to facilitate community-driven development. Availability and Implementation OCMV is available at https://github.com/ShangWangLab/OpenChronoMorphViewer for Windows, Linux, and macOS. Supporting tutorials, documentation, and installation instructions can be found in the supplementary information. Our modified Fiji I/O plugin for up to 5D NRRD file conversion is available at https://github.com/afaubert/IO. Supplementary information Supplementary Figures, Tables, Videos, and Notes are available at Bioinformatics online.

Exploring 3D Modeling for Enhanced Visualization and Analysis of Pediatric Pelvic Ring Fractures: A Pilot Study

Introduction Pediatric pelvic fractures are high energy injuries that present substantial treatment challenges due to their propensity for plastic deformation, complex fracture patterns, and lack of consensus on surgical indications. There is potential for negative long-term outcomes in patients with residual deformity, but our understanding of acceptable alignment is limited. Plain radiographs and computed tomography (CT) are frequently employed for pre- and post-operative assessment of injury patterns, but are limited to two-dimensional analysis. Three dimensional (3D) modeling offers an opportunity for improved assessment of pelvic anatomy. Methods Two pediatric patients who sustained unstable pelvic ring injuries requiring operative fixation, with both pre- and post-operative CT scans, were identified from an existing IRB-approved trauma database. Utilizing 3D Slicer, 3D segmentations were rendered from DICOM data and specific landmarks were identified to calculate deformity pre- and post-operatively including displacement at the sacroiliac (SI) joints, pubic symphysis, and iliac crest. A modified Keshishyan deformity index was also calculated. Mesh overlay was performed using Autodesk Maya in x, y and z planes to provide visual representation of post-operative changes. Results 3D renderings successfully illustrated complex pelvic fractures in a 9 year old female and 12 year old. Measurements made on 3D models showed similar improvements in diastasis and symmetry to measurements made in standard 2D imaging, but with generally smaller magnitudes. Graphic overlap allow for sophisticated representation of improvement in deformity. Measurements on plain radiographs and 2D CT differed from 3D measurements by anywhere from 0.1 mm to 6.7 mm. Conclusion This proof-of-concept pilot study demonstrates that 3D renderings enhance the ability to measure displacement and deformity in pediatric pelvic trauma for both pre-operative planning and post-operative evaluation. Future applications of this technology include evaluation of pelvic deformity remodeling after trauma in both operative and non-operatively treated pediatric patients.

Quantitative Evaluation of Improvement of Tear Trough With a Non-Cross-Linked Sodium Hyaluronic Compound: A Three-Dimensional and MRI Analysis.

Tear trough deformities are indicative of aging, progressively manifesting as pronounced infraorbital hollows. Although hyaluronic acid (HA) dermal fillers are favored for their safety and efficacy, quantifiable evidence of their effect of action still remains insufficient. To investigate the efficacy of non-cross-linked HA compound in tear trough enhancement. Twenty-one subjects with moderate to severe infraorbital hollows underwent a single treatment including subcutaneous to supraperiosteal injections of a non-cross-linked HA compound. The effectiveness was assessed through Allergan Infraorbital Hollows Scale (AIHS), satisfaction rates, the Global Aesthetic Improvement Scale (GAIS), and quantitative analyses via Canfield VECTRA 3D imaging and MRI measurements during the 6-month follow-up. A total of 0.67 ± 0.31 and 0.77 ± 0.42 mL of the non-cross-linked HA compound were injected into the left and right suborbital regions, respectively. Substantial improvements in AIHS were observed in 85.71% and 80.95% of subjects at 6 months posttreatment, as assessed by blinded evaluators and treating investigators, respectively. 3D imaging showed volume increases of 0.87 ± 0.32 (left) and 0.99 ± 0.45 mL (right). MRI analysis provided closely aligned results to the 3D analysis. The GAIS and satisfaction rate were 85.71%, 90.48% for treating investigators and were 90.48% and 95.24% for subjects 6 months posttreatment. The non-cross-linked HA filler demonstrated a safe and efficacious profile for the correction of infraorbital hollows with significant patient satisfaction and sustained outcomes up to 6 months posttreatment. These results support their clinical value in periorbital rejuvenation and reduction of infraorbital hollowing.

STNeRF: symmetric triplane neural radiance fields for novel view synthesis from single-view vehicle images

This paper presents STNeRF, a method for synthesizing novel views of vehicles from single-view 2D images without the need for 3D ground truth data, such as point clouds, depth maps, CAD models, etc., as prior knowledge. A significant challenge in this task arises from the characteristics of CNNs and the utilization of local features can lead to a flattened representation of the synthesized image when training and validation with images from a single viewpoint. Many current methodologies tend to overlook local features and rely on global features throughout the entire reconstruction process, potentially resulting in the loss of fine-grained details in the synthesized image. To tackle this issue, we introduce Symmetric Triplane Neural Radiance Fields (STNeRF). STNeRF employs a triplane feature extractor with spatially aware convolution to extend 2D image features into 3D. This decouples the appearance component, which includes local features, and the shape component, which consists of global features, and utilizes them to construct a neural radiance field. These neural priors are then employed for rendering novel views. Furthermore, STNeRF leverages the symmetric properties of vehicles to liberate the appearance component from reliance on the original viewpoint and to align it with the symmetry of the target space, thereby enhancing the neural radiance field network’s ability to represent the invisible regions. The qualitative and quantitative evaluations demonstrate that STNeRF outperforms existing solutions in terms of both geometry and appearance reconstruction. More supplementary materials and the implementation code are available for access at the following link: https://github.com/ll594282475/STNeRF.

White Matter Fiber Bundle Alterations Correlate with Gait and Cognitive Impairments in Parkinson’s Disease based on HARDI Data

Background: The neuroanatomical basis of white matter fiber tracts in gait impairments in individuals suffering from Parkinson’s Disease (PD) is unclear. Methods: Twenty-four individuals living with PD and 29 Healthy Controls (HCs) were included. For each participant, two-shell High Angular Resolution Diffusion Imaging (HARDI) and high-resolution 3D structural images were acquired using the 3T MRI. Diffusion-weighted data preprocessing was performed using the orientation distribution function to trace the main fiber tracts in PD individuals. Clinical characteristics between the two groups were compared, and the correlation between the FA value and behavioral data was analyzed. Quantitative gait and clinical parameters were recorded in PD at ON and OFF states, respectively. Results: The mean tract-specific FA values of the right Cingulum Cingulate (rCC) were statistically different between the PD group and the HC group (p =0.047). The FA value of 34-58 equidistant nodes in rCC was positively correlated with Mini-Mental State Examination (MMSE) (r=0.527, p=0.024), Berg Balance Scale (BBS)-OFF (r=0.480, p =0.040), and BBS-ON (r=0.528, p =0.024) scores, while it was negatively correlated with the MDS-UPDRS-III-ON score (r=-0.502, p =0.030). Regarding the gait analysis, the FA value was significantly correlated with velocity, cadence, and stride time of the pace and rhythm domains in both ‘ON’ and ‘OFF’ states, respectively (p<0.05). Conclusion: This study served as an initial exploration to establish that HARDI sequences could be employed as a robust tool for analyzing microstructural alterations in white matter fiber bundles among PD patients, although the sample size was small. We confirmed microstructural integrity impairment of rCC to be significantly associated with both gait and cognitive deficits in patients with PD. Early detection of microstructural changes in rCC and targeted treatment can help improve behavioral disorders. In the future, we intend to further integrate multimodal data with assessments of patient behavior both prior to and following intervention. We will validate our findings within an independent cohort to monitor disease progression and evaluate the efficacy of therapeutic interventions.

Operando X-Ray Tomoscopy of Laser Beam Welding.

The phenomena occurring in a weld seam during advancement of a laser beam over a metallic component are still under dispute. The occurrence and evolution of porosity and the occasional blowout of melt need to be understood. Here, a recently developed X-ray tomoscopy setup is applied, providing one hundred 3D images per second to capture the temporal evolution of the melt pool in an AlSi9Cu3(Fe) die-casting while a laser beam advances. The number of pores, their size, shape and distribution are quantified with 10 ms time resolution and reflect a complex dynamic pattern. Apart from conventional welding, a variant involving a dynamic beam modulation superimposed onto the linear motion is studied. Reductions of porosity and surface roughness are observed and explained by increased pore mobility and stepwise degassing as the beam repeatedly cuts through pores. The keyhole formed in the melt pool integrated over 10 ms is represented in3D.

PDO screw threads versus fractional CO2 laser, after subcision, for atrophic acne scar remodelling: a randomized group-based comparative study.

Atrophic acne scars present a significant therapeutic challenge. While subcision with various adjunctive treatments, including fractional CO₂ lasers and polydioxanone (PDO) threads, has been employed for scar remodeling, comparative evidence on their efficacy remains limited. This study aims to compare the clinical efficacy and patient satisfaction between subcision with fractional CO₂ laser and subcision with PDO screw threads in managing atrophic acne scars. A randomized, controlled study was conducted on 40 patients with atrophic acne scars. Patients were assigned to two groups: Group A underwent subcision with PDO screw threads insertion, and Group B underwent subcision followed by fractional CO₂ laser resurfacing. Outcomes were assessed using the Goodman and Baron grading system, Antera 3D imaging, and patient satisfaction. Significant scar improvement was observed in both groups, with Group B demonstrating greater reduction in scar severity and depth (p = 0.022). While Group A showed faster recovery and lower downtime, Group B yielded higher patient satisfaction scores. Subcision combined with fractional CO₂ laser resulted in superior long-term scar improvement, although PDO threads provided quicker recovery. The choice of treatment may depend on patient preferences and scar characteristics.

Three-Dimensional Electrical Impedance Imaging During Spontaneous Breathing Trials in Patients With Acute Hypoxic Respiratory Failure: A Pilot Study.

The purpose of this work is to evaluate the feasibility of lung imaging using 3D electrical impedance tomography (EIT) during spontaneous breathing trials (SBTs) in patients with acute hypoxic respiratory failure. EIT is a noninvasive, nonionizing, real-time functional imaging technique, suitable for bedside monitoring in critically ill patients. EIT data were collected in 24 mechanically ventilated patients immediately preceding and during a SBT on two rows of 16 electrodes using a simultaneous multicurrent source EIT system for 3D imaging. Dynamic 3D EIT images of conductivity were computed, as well as the EIT-derived rapid shallow breathing index, regional ventilation delay, global inhomogeneity index, and time traces of tidal volumes. 3D reconstructions and derived measures demonstrated inhomogeneity in ventilation distribution within patients. We conclude that 3D EIT images can provide information regarding ventilatory heterogeneity across the lung and may be useful in guiding ventilator management.

Computed Tomography Evaluation of Morphological Types of Femoral Trochlear Dysplasia in Small-Breed Dogs—A Retrospective Study

Abnormal trochlear morphology is one of the most important factors for patellar luxation occurrence in dogs, yet no studies have investigated its prevalence in the general population. This retrospective computed tomography study was designed to evaluate the trochlear groove morphology in four small dog breeds and the prevalence of trochlear dysplasia types according to Déjour’s classification depending on the breed, sex, and medial patellar luxation (MPL) presence and grade. A total of 174 joints (68 healthy, 96 grade II MPL, and 10 grade III MPL) from Mini-Pinschers, Yorkshire Terriers, Pomeranians, and Chihuahuas were included in the study. The morphological type of trochlear dysplasia (TD) was evaluated on axial scans and 3D reconstruction images according to the four-type classification of Déjour, sulcus angle, trochlear depth, and lateral/medial inclination angles. Of all 174 joints, 140 had trochlear dysplasia—all joints with MPL (n = 106) and 50% of healthy joints (n = 34). The classification of Déjour for trochlear dysplasia types (A, B, C, and D) corresponds to the morphology of the femoral trochlea in the studied small breeds of dogs. The results demonstrated three types of trochlear dysplasia according to Déjour: most commonly, type A, followed by type C, and most infrequently, type D. The Déjour type B was an incidental finding. The large proportion of clinically healthy joints with TD (50%) emphasises the significance of early trochlear morphology evaluation for the orthopaedical health of dogs from susceptible breeds, especially in female breeders.

Pulmonary Atresia with Ventricular Septal Defect without Major Aorto-Pulmonary Collateral Arteries: Echocardiography and the Role of Computed Tomography and Magnetic Resonance Imaging.

Pulmonary atresia with ventricular septal defect (PA-VSD) is usually diagnosed by transthoracic or fetal echocardiography, with the prenatal diagnosis being feasible and accurate if fetal cardiology services are available. The limitations of transthoracic echocardiography (TTE) in the evaluation of PA-VSD include the complete evaluation of the pulmonary arteries and patent ductus arteriosus, quantitative evaluation of the right ventricle size and function, and delineation of associated cardiac anomalies such as coronary artery anomalies, anomalies of systemic or pulmonary venous return, and complex arch anomalies. Echocardiography also has limitations in evaluating hemodynamics such as flow volumes, shunts, and regurgitant fraction. Despite these limitations, TTE remains the most widely available and the most cost-effective cardiac imaging modality for patients with PA-VSD and its accuracy in evaluating the sources of pulmonary blood flow and in selecting patients for systemic-to-pulmonary artery shunt palliation is well established. Cardiac computed tomography (CT) can answer many of the questions not answered by TTE, including demonstrating the PA anatomy and defining coronary artery, systemic and pulmonary venous, and aortic arch anatomies. The short acquisition time allows for the study to be performed without sedation/anesthesia in most patients. Cardiac CT is also useful in defining postoperative anatomy when there are non-magnetic resonance imaging (MRI) compatible devices or even MRI-compatible devices that cause a significant MRI artifact. Cardiac MRI/magnetic resonance angiography has emerged as an ideal modality to evaluate patients with PA-VSD as it allows for anatomic, functional, and hemodynamic assessment without exposure to ionizing radiation or iodinated contrast material. Both cardiac CT and cardiac MRI can be used to generate 3D imaging of the heart.

PEEK stress-shielding with artificial bone for dental implants.

Dental titanium implants and their surface modifications markedly improve implant biocompatibility. However, studies evaluating the mechanical biocompatibility of implants are scarce. In particular, the analysis of mechanical biocompatibility deficiencies leading to stress shield-induced bone resorption. Recently, we focused on using PEEK as a dental material. This study explored the hypothesis that PEEK implants improve the stress shielding of titanium. In this study, artificial bone surfaces were examined to measure strains on the artificial bone surface under compressive loading with the implants in place. Additionally, 3D image analysis of the fracture state inside the bone tissue was performed using micro-CT (µCT). This hypothesis was supported by µCT imaging analysis of bone tissue changes under stress, which revealed that PEEK implants transfer greater loads than titanium implants. µCT imaging and statistical analysis showed that bone porosity had little effect on stress shielding.

Automatic 4D mitral valve segmentation from transesophageal echocardiography: a semi-supervised learning approach.

Performing automatic and standardized 4D TEE segmentation and mitral valve analysis is challenging due to the limitations of echocardiography and the scarcity of manually annotated 4D images. This work proposes a semi-supervised training strategy using pseudo labelling for MV segmentation in 4D TEE; it employs a Teacher-Student framework to ensure reliable pseudo-label generation. 120 4D TEE recordings from 60 candidates for MV repair are used. The Teacher model, an ensemble of three convolutional neural networks, is trained on end-systole and end-diastole frames and is used to generate MV pseudo-segmentations on intermediate frames of the cardiac cycle. The pseudo-annotated frames augment the Student model's training set, improving segmentation accuracy and temporal consistency. The Student outperforms individual Teachers, achieving a Dice score of 0.82, an average surface distance of 0.37 mm, and a 95% Hausdorff distance of 1.72 mm for MV leaflets. The Student model demonstrates reliable frame-by-frame MV segmentation, accurately capturing leaflet morphology and dynamics throughout the cardiac cycle, with a significant reduction in inference time compared to the ensemble. This approach greatly reduces manual annotation workload and ensures reliable, repeatable, and time-efficient MV analysis. Our method holds strong potential to enhance the precision and efficiency of MV diagnostics and treatment planning in clinical settings.

A Dual-Path Neural Network for High-Impedance Fault Detection

High-impedance fault detection poses significant challenges for distribution network maintenance and operation. We propose a dual-path neural network for high-impedance fault detection. To enhance feature extraction, we use a Gramian Angular Field algorithm to transform 1D zero-sequence voltage signals into 2D images. Our dual-branch network simultaneously processes both representations: the CNN extracts spatial features from the transformed images, while the GRU captures temporal features from the raw signals. To optimize model performance, we integrate the Crested Porcupine Optimizer (CPO) algorithm for the adaptive optimization of key network hyperparameters. The experimental results demonstrate that our method achieves a 99.70% recognition accuracy on a dataset comprising high-impedance faults, capacitor switching, and load connections. Furthermore, it maintains robust performance under various test conditions, including different noise levels and network topology changes.