Three-dimensional Medical Images Research Articles

Enhancing cerebral arteriovenous malformation analysis: Development and application of patient-specific lumped parameter models based on 3D imaging data

Objectives:Cerebral arteriovenous malformations (AVMs) present complex neurovascular challenges, characterized by direct arteriovenous connections that disrupt normal brain blood flow dynamics. Traditional lumped parameter models (LPMs) offer a simplified angioarchitectural representation of AVMs, yet often fail to capture the intricate structure within the AVM nidus. This research aims at refining our understanding of AVM hemodynamics through the development of patient-specific LPMs utilizing three-dimensional (3D) medical imaging data for enhanced structural fidelity. Methods:This study commenced with the meticulous delineation of AVM vascular architecture using threshold segmentation and skeletonization techniques. The AVM nidus’s core structure was outlined, facilitating the extraction of vessel connections and the formation of a detailed fistulous vascular tree model. Sampling points, spatially distributed and derived from the pixel intensity in imaging data, guided the construction of a complex plexiform tree within the nidus by generating smaller Y-shaped vascular formations. This model was then integrated with an electrical analog model to enable precise numerical simulations of cerebral hemodynamics with AVMs. Results:The study successfully generated two distinct patient-specific AVM networks, mirroring the unique structural and morphological characteristics of the AVMs as captured in medical imaging. The models effectively represented the intricate fistulous and plexiform vessel structures within the nidus. Numerical analysis of these models revealed that AVMs induce a blood shunt effect, thereby diminishing blood perfusion to adjacent brain tissues. Conclusion:This investigation enhances the theoretical framework for AVM research by constructing patient-specific LPMs that accurately reflect the true vascular structures of AVMs. These models offer profound insights into the hemodynamic behaviors of AVMs, including their impact on cerebral circulation and the blood steal phenomenon. Further incorporation of clinical data into these models holds the promise of deepening the theoretical comprehension of AVMs and fostering advancements in the diagnosis and treatment of AVMs.

Volumetric analysis of acute uncomplicated type B aortic dissection using an automated deep learning aortic zone segmentation model

BackgroundMachine learning techniques have shown excellent performance in three-dimensional medical image analysis, but have not been applied to acute uncomplicated type B aortic dissection (auTBAD) using Society for Vascular Surgery (SVS) and Society of Thoracic Surgeons (STS)-defined aortic zones. The purpose of this study was to establish a trained, automatic machine learning aortic zone segmentation model to facilitate performance of an aortic zone volumetric comparison between patients with auTBAD based on the rate of aortic growth. MethodsPatients with auTBAD and serial imaging were identified. For each patient, imaging characteristics from two computed tomography (CT) scans were analyzed: (1) the baseline CT angiography (CTA) at the index admission and (2) either the most recent surveillance CTA or the most recent CTA before an aortic intervention. Patients were stratified into two comparative groups based on aortic growth: rapid growth (diameter increase of ≥5 mm/year) and no or slow growth (diameter increase of <5 mm/year). Deidentified images were imported into an open source software package for medical image analysis and images were annotated based on SVS/STS criteria for aortic zones. Our model was trained using four-fold cross-validation. The segmentation output was used to calculate aortic zone volumes from each imaging study. ResultsOf 59 patients identified for inclusion, rapid growth was observed in 33 patients (56%) and no or slow growth was observed in 26 patients (44%). There were no differences in baseline demographics, comorbidities, admission mean arterial pressure, number of discharge antihypertensives, or high-risk imaging characteristics between groups (P > .05 for all). Median duration between baseline and interval CT was 1.07 years (interquartile range [IQR], 0.38-2.57). Postdischarge aortic intervention was performed in 13 patients (22%) at a mean of 1.5 ± 1.2 years, with no difference between the groups (P > .05). Among all patients, the largest relative percent increases in zone volumes over time were found in zone 4 (13.9%; IQR, −6.82 to 35.1) and zone 5 (13.4%; IQR, −7.78 to 37.9). There were no differences in baseline zone volumes between groups (P > .05 for all). The average Dice coefficient, a performance measure of the model output, was 0.73. Performance was best in zone 5 (0.84) and zone 9 (0.91). ConclusionsWe describe an automatic deep learning segmentation model incorporating SVS-defined aortic zones. The open source, trained model demonstrates concordance to the manually segmented aortas with the strongest performance in zones 5 and 9, providing a framework for further clinical applications. In our limited sample, there were no differences in baseline aortic zone volumes between patients with rapid growth and patients with no or slow growth.

Sailfish optimizer based CLAHE with U-NET for MRI brain tumour segmentation

Semantic brain tumour segmentation is a significant contribution to medical image processing. This research is helpful for doctors making diagnoses and determining the severity of lesions. In recent years, convolutional neural networks have sometimes been known as self-remarkable representations in computer vision tasks. In three-dimensional medical image processing, CNN with architecture and skip connections has become increasingly popular. However, they are inadequate at both acquiring nearby and far-off semantic knowledge. However, the transformer's use of a vision. This accomplishment was made possible by the global information modelling capabilities of the transformer. When tryingto perform challenging prediction tasks, such as segmenting 3D medical pictures, it is vital to consider local and global features. In this study, we propose a hypothesis that predicts pre-processing steps that enhance the image quality and significantly impact accuracy with any statistical image classification method. We used a better picture enhancement methodology to back up our idea. This method has three unique stages: a median filter to eliminate noise, a histogram equalisation technique to boost contrast, and a grayscale-to-RGB conversion. The clipping limit of contrast-limited Adaptive Histogram Equalisation (CLAHE)is calculated by the Sailfish Optimization Algorithm (SOA) to achieve optimal picture enhancement. This study classifies semantically segmenting 3D brain tumours as a sequence-to-sequence prediction problem. U-net is proposed in the article as a novel method for segmenting 3D medical images; it incorporates a structure. A new technique for segmenting 3D medical images, U-net, is emerging. We trained and assessed our method on the BRATS 2019 dataset, which was found to outperform state-of-the-art algorithms and be competitive under similar settings.

EMBRYOLOGICAL PREREQUISITES FOR HUMAN CRANIAL DEVELOPMENTAL DEFFECTS

The purpose of the work is to find out the embryological prerequisites and the time of the possible occurrence of malformations and congenital deformations of the human skull.Results. The growth and enlargement of the neurocranium is largely determined by the growth of the brain due to their close syntopy. By the end of the second year, the bones are connected by sutures. The neurocranium is embryologically divided into the vault, which is formed by membranous ossification, and the basis cranii, the bones of which are formed by cartilaginous osteogenesis. The initial development of the neurocranium depends on the formation of the brain and its surrounding membranes. The source of the germ of the outer layer is the mesoderm and ectomesenchyme of the neural crest, soon it is divided into the inner layer – endomeninx and outer – ectomeninx. The latter is further divided into the outer osteogenic layer, in which the ossification centers form the bones of the skull, and the inner layer of the dura mater. The centers of ossification form the frontal, parietal, occipital and temporal bones, and the intermediate areas form fibrous sutures and fontanelles. Usually, the sutures are joined at the end of the second year of life. Conclusions. Congenital deformities of the cerebral part of the skull are morphological manifestations of numerous hereditary syndromes and/or metabolic, teratogenic effects, lysosomal storage diseases, brain malformations, unfavorable factors during the fetal period of development: restriction of the intrauterine environment, weak muscle tone, torticollis, clavicle fracture, cervical spine anomalies, multiple pregnancy and bone mineralization disorders. For the differential diagnosis of congenital defects of the skull, along with the use of diagnostic three-dimensional medical imaging, it is necessary to carry out genetic screening.

Numerical Simulations of the Epley Maneuver With Clinical Implications.

Canalith repositioning procedures to treat benign paroxysmal positional vertigo are often applied following standardized criteria, without considering the possible anatomical singularities of the membranous labyrinth for each individual. As a result, certain patients may become refractory to the treatment due to significant deviations from the ideal membranous labyrinth, that was considered when the maneuvers were designed. This study aims to understand the dynamics of the endolymphatic fluid and otoconia, within the membranous labyrinth geometry, which may contribute to the ineffectiveness of the Epley maneuver. Simultaneously, the study seeks to explore methods to avoid or reduce treatment failure. We conducted a study on the Epley maneuver using numerical simulations based on a three-dimensional medical image reconstruction of the human left membranous labyrinth. A high-quality micro-computed tomography of a human temporal bone specimen was utilized for the image reconstruction, and a mathematical model for the endolymphatic fluid was developed and coupled with a spherical particle model representing otoconia inside the fluid. This allowed us to measure the position and time of each particle throughout all the steps of the maneuver, using equations that describe the physics behind benign paroxysmal positional vertigo. Numerical simulations of the standard Epley maneuver applied to this membranous labyrinth model yielded unsatisfactory results, as otoconia do not reach the frontside of the utricle, which in this study is used as the measure of success. The resting times between subsequent steps indicated that longer intervals are required for smaller otoconia. Using different angles of rotation can prevent otoconia from entering the superior semicircular canal or the posterior ampulla. Steps 3, 4, and 5 exhibited a heightened susceptibility to failure, as otoconia could be accidentally displaced into these regions. We demonstrate that modifying the Epley maneuver based on the numerical results obtained in the membranous labyrinth of the human specimen under study can have a significant effect on the success or failure of the treatment. The use of numerical simulations appears to be a useful tool for future canalith repositioning procedures that aim to personalize the treatment by modifying the rotation planes currently defined as the standard criteria.

Towards the Emergence of the Medical Metaverse: A Pilot Study on Shared Virtual Reality for Orthognathic–Surgical Planning

Three-dimensional (3D) medical images are used for diagnosis and in surgical operation planning. Computer-assisted surgical simulations (CASS) are essential for complex surgical procedures that are often performed in an interdisciplinary manner. Traditionally, the participants study the designs on the same display. In 3D virtual reality (VR) environments, the planner is wearing a head-mounted display (HMD). The designs can be then examined in VR by other persons wearing HMDs, which is a practical use case for the medical metaverse. A multi-user VR environment was built for the planning of an orthognathic–surgical (correction of facial skeleton) operation. Four domain experts (oral and maxillofacial radiologists) experimented with the pilot system and found it useful. It enabled easier observation of the model and a better understanding of the structures. There was a voice connection and co-operation during the procedure was natural. The planning task is complex, leading to a certain level of complexity in the user interface.

AI-powered Hyperrealism: Next Step in Cinematic Rendering?

Background Recent advancements in artificial intelligence (AI)-powered image generation present opportunities to enhance three-dimensional medical images. Diffusion, an iterative denoising process, represents the standard of many of the current tools used for this purpose. Purpose To demonstrate the current capabilities of diffusion technology by using Midjourney, version 5.2, a text-to-image generative AI tool, and present a practical guide for its use. Materials and Methods This exploratory study investigates the principles, parameters, and prompt engineering techniques for generating images focusing on Midjourney from July 27 to August 3, 2023. Step-by-step instructions show the innate capability of this technology in creating realistic medical images. Results Thirty images were selected, including eye, skin, and vascular aneurysm images. Varying prompt phrasing and weighting techniques allowed for the customization of output image characteristics. Although the details of Midjourney's model training are confidential, it is estimated that it was trained on at least hundreds of millions of images from the web. Anatomic fidelity was not always maintained because the training data set is not necessarily based on accurate medical images. There are shortcomings in this nascent technology regarding its ability to create entities such as digits of the hand or precise text. Conclusion AI image generation has the potential to improve three-dimensional medical images for certain applications through added visual detail and appeal but ongoing collaboration is needed between radiologists and AI developers due to the overreliance on art and photography in the training data, which may result in inaccurate anatomic results. Moreover, the evolving landscape of ethical discussions and copyright stipulations warrants close attention. © RSNA, 2024 Supplemental material is available for this article.

Residual learning for segmentation of the medical images in healthcare

Medical workers can assess disease progression and create expedient treatment plans with the help of automated and accurate 3Dsegmentation of medical images. DCNNs (Deep convolution neural networks) have been widely used in this work, but their accuracy still needs to be increased, mostly due to their insufficient understanding of 3D environments. This study proposed three dimensional residual networks, ResUNet++, for precise segmentations of three-dimensional medical images where encoders, segmentation decoders, and context residual decoders are used. Two decoders are connected at scale utilizing context attention maps and context residual, the former explicitly learns inter-slice context data and the latter utilizes contexts as attention to increase segmentation accuracy. This model was assessed by using MICCAI 2018 BraTS dataset and, the Pancreas-CT dataset. The BrasTS and Pancreas-CT dataset scales were compared in terms of ET, WT, TC. Moreover, the proposed model was compared with/without boundary loss and validation dice score. The outcomes not only show how effective the suggested 3D residual learning approach is, but also show that the suggested ResUNet++ offers better accuracy compared to six of the top-ranking techniques used for segmenting tumors in the brain.

ASSOCIATION OF MASSETER MUSCLE VOLUME WITH RISK OF OSTEOPOROSIS IN OLDER ADULTS: THE BUNKYO HEALTH STUDY

Abstract Abstract The relationship between bone mineral density (BMD) and oral function has been previously studied. However, the relationship between the masseter muscle, which plays an important role in oral function, and BMD and osteoporosis in older adults has not been investigated. This study investigated the association between masseter muscle volume (MMV) and old age in relation to osteoporosis prevalence and hip joint BMD. Body composition and BMD of 1534 people aged 65–84 years living in Bunkyo-ku, Tokyo, were measured using dual-energy X-ray absorptiometry. MMV was measured using a three-dimensional medical image analysis system. Participants were divided into five groups based on MMV by 20% for each gender (Quintile 1–5: Q1–5, Q1&amp;lt; Q5). Logistic regression models were employed to estimate multivariable-adjusted odds ratios (ORs) for osteoporosis determined by T-score (&amp;lt;−2.5 standard deviation) using the Q5 reference. Subsequently, analysis of covariance was used to investigate the relationship between MMV and BMD. Adjustment for age, body mass index, smoking history, alcohol intake, calcium intake, 25(OH)D, diabetes, and osteoporosis drugs or estrogen. In men, Q1 had higher hip osteoporosis prevalence than Q5 (OR 11.65; 95% confidence interval, 1.48–1.62, p=0.020). Further, hip BMD was significantly higher in Q5 than in Q1 (p=0.001). In women, MMV and old age was not associated with osteoporosis prevalence. However, hip BMD was significantly higher in Q3,4 than in Q1 (p=0.003). Older adults with lower MMV had lower BMD in the hip joint. Additionally, men had higher odds of osteoporosis of the hip joint.

Modern Image-Guided Surgery: A Narrative Review of Medical Image Processing and Visualization.

Medical image analysis forms the basis of image-guided surgery (IGS) and many of its fundamental tasks. Driven by the growing number of medical imaging modalities, the research community of medical imaging has developed methods and achieved functionality breakthroughs. However, with the overwhelming pool of information in the literature, it has become increasingly challenging for researchers to extract context-relevant information for specific applications, especially when many widely used methods exist in a variety of versions optimized for their respective application domains. By being further equipped with sophisticated three-dimensional (3D) medical image visualization and digital reality technology, medical experts could enhance their performance capabilities in IGS by multiple folds. The goal of this narrative review is to organize the key components of IGS in the aspects of medical image processing and visualization with a new perspective and insights. The literature search was conducted using mainstream academic search engines with a combination of keywords relevant to the field up until mid-2022. This survey systemically summarizes the basic, mainstream, and state-of-the-art medical image processing methods as well as how visualization technology like augmented/mixed/virtual reality (AR/MR/VR) are enhancing performance in IGS. Further, we hope that this survey will shed some light on the future of IGS in the face of challenges and opportunities for the research directions of medical image processing and visualization.

Full 3-D modulation transfer function estimation of tomosynthesis system using modified Richardson-Lucy deconvolution.

Tomosynthesis systems are three-dimensional (3-D) medical imaging devices that operate over limited acquisition angles using low radiation dosages. To measure the spatial resolution performance of a tomosynthesis system, the modulation transfer function (MTF) is widely used as a quantitative evaluationmetric. We previously introduced a method to estimate the full 3-D MTF of a cone-beam computed tomography system using two-dimensional (2-D) Richardson-Lucy (RL) deconvolution with Tikhonov-Miller regularization. However, this method can not be applied directly to estimate the 3-D MTF of a tomosynthesis system, since the unique artifacts (i.e., shadow artifacts, spreading tails, directional blurring, and high-level noise) of the system produce several errors that lower the estimation performance. Varying positions of the negative pixels due to shadow artifacts and spreading tails cause inconsistent deconvolution performances at each of the directional projections, and the severe noise in the reconstructed images cause noise amplification during estimation. This work proposes several modifications to the previous method to resolve the inconsistent performance and noise amplification errors to increase the full 3-D MTF estimationaccuracy. Three modifications were introduced to the 2-D RL deconvolution to prevent estimation errors and improve MTF estimation performance: non-negativity relaxation function, cost function to terminate the iterative process of RL deconvolution, and regularization strength for noise control. To validate the effectiveness of the proposed modifications, we reconstructed sphere phantoms from simulation and experimental tomosynthesis studies in the iso-center and offset-center positions as well as estimated the full 3-D MTFs using the previous and proposed methods. We compared the 3-D render images, central plane images, and center profiles of the estimated 3-D MTFs and calculated the full widths at half and tenth maximum for quantitativeevaluation. The previous method cannot estimate the full 3-D MTF of a tomosynthesis system; its inaccurate negative pixel relaxation produces circular-shaped errors, and the mean squared error based simple cost function for termination causes inconsistent estimation at each directional projection to diminish the clear edges of the low-frequency drop and missing sample regions. Noise amplification from lack of noise regularization is also seen in the previous method results. Compared to the previous method, the proposed method shows superior estimation performance at reducing errors in both the simulation and experimental studies regardless of object position. The proposed method preserves the low-frequency drop, missing sample regions from the limited acquisition angles, and missing cone region from the offset-center position; the estimated MTFs also show FWHM and FWTM values close to those of the ideal MTFs than with the previousmethod. This work presents a method to estimate the full 3-D MTF of a tomosynthesis system. The proposed modifications prevent circular-shaped errors and noise amplification due to the geometry for limited acquisition angles and high noise levels. Compared to our previous method, the proposed scheme show better performance for estimating the 3-D MTF of the tomosynthesissystem.

Kernel-weighted contribution: a method of visual attribution for 3D deep learning segmentation in medical imaging.

Explaining deep learning model decisions, especially those for medical image segmentation, is a critical step toward the understanding and validation that will enable these powerful tools to see more widespread adoption in healthcare. We introduce kernel-weighted contribution, a visual explanation method for three-dimensional medical image segmentation models that produces accurate and interpretable explanations. Unlike previous attribution methods, kernel-weighted contribution is explicitly designed for medical image segmentation models and assesses feature importance using the relative contribution of each considered activation map to the predicted segmentation. We evaluate our method on a synthetic dataset that provides complete knowledge over input features and a comprehensive explanation quality metric using this ground truth. Our method and three other prevalent attribution methods were applied to five different model layer combinations to explain segmentation predictions for 100 test samples and compared using this metric. Kernel-weighted contribution produced superior explanations of obtained image segmentations when applied to both encoder and decoder sections of a trained model as compared to other layer combinations (). In between-method comparisons, kernel-weighted contribution produced superior explanations compared with other methods using the same model layers in four of five experiments () and showed equivalently superior performance to GradCAM++ when only using non-transpose convolution layers of the model decoder (). The reported method produced explanations of superior quality uniquely suited to fully utilize the specific architectural considerations present in image and especially medical image segmentation models. Both the synthetic dataset and implementation of our method are available to the research community.

The evaluation of vasculature in post-mortem angio-computed tomography for anatomy research purposes: method description based on celiac trunk analysis.

Anatomical research based on deceased body specimens is a time-consuming process that requires a great deal of skill and time to perform correctly. Three-dimensional medical image analysis is an excellent tool for anatomic evaluation, but it often includes patients with comorbidities in the study group, which can skew the results. The purpose of this study was to develop and evaluate methods for anatomic research based on postmortem contrast-enhanced computed tomography angiography 3D reconstruction of the celiac trunk. Postmortem contrast-enhanced computed tomography angiography of 105 (28.6% female, age 50.8±18.7) decedents without abdominal trauma or tumor was analyzed. The abdominal portion of the aorta and the celiac trunk with its branches were reconstructed and evaluated. The type of celiac trunk was evaluated. The results were analyzed. The celiac trunk, splenic artery, and common hepatic artery were visualized in all cases. The left gastric artery was visible in 97.1% of cases. The dorsal pancreatic artery was visualized in 61.0% of cases. The most common type of celiac trunk was 1 (88.6%), and the rarest types were 2, 3, and 6 (1.0%). We observed 4 morphologies of the truncus celiacus that did not fit the classification presented previously. This study has demonstrated that three-dimensional reconstruction of postmortem contrast-enhanced computed tomography is an excellent tool for performing accurate morphometric analyzes for anatomic research purposes. This method can serve as a source for anatomic studies in the healthy population.

Semi-automatic Framework for Voxel Human Deformation Modeling.

With the advancement of computer and medical imaging technologies, a number of high-resolution, voxel-based, full-body human anatomical models have been developed for medical education, industrial design, and physics simulation studies. However, these models are limited in many applications because they are often only in an upstanding posture. To quickly develop multi-pose human models for different applications. A semi-automatic framework for voxel deformation is proposed in the study. This paper describes a framework for human pose deformation based on three-dimensional (3D) medical images. The voxel model is first converted into a surface model using a surface reconstruction algorithm. Second, a deformation skeleton based on human bones is defined, and the surface model is bound to the skeleton. The bone Glow algorithm is used to assign weights to the surface vertices. Then, the model is deformed to the target posture by using the Smoothed Rotation Enhanced As-Rigid-As-Possible (SR-ARAP) algorithm. Finally, the volume-filling algorithm is applied to refill the tissues into the deformed surface model. The proposed framework is used to deform two standing human models, and the sitting and running models are developed. The results show that the framework can successfully develop the target pose. When compared to the results of the As-Rigid-As-Possible algorithm, SR-ARAP preserves local tissues better. The study proposes a frame for voxel human model deformation and improves the local tissue integrity during deformation.

SUNet++: A Deep Network with Channel Attention for Small-Scale Object Segmentation on 3D Medical Images

As a deep learning network with an encoder-decoder architecture, UNet and its series of improved versions have been widely used in medical image segmentation with great applications. However, when used to segment targets in 3D medical images such as magnetic resonance imaging (MRI), computed tomography (CT), these models do not model the relevance of images in vertical space, resulting in poor accurate analysis of consecutive slices of the same patient. On the other hand, the large amount of detail lost during the encoding process makes these models incapable of segmenting small-scale tumor targets. Aiming at the scene of small-scale target segmentation in 3D medical images, a fully new neural network model SUNet++ is proposed on the basis of UNet and UNet++. SUNet++ improves the existing models mainly in three aspects: 1) the modeling strategy of slice superposition is used to thoroughly excavate the three dimensional information of the data; 2) by adding an attention mechanism during the decoding process, small scale targets in the picture are retained and amplified; 3) in the up-sampling process, the transposed convolution operation is used to further enhance the effect of the model. In order to verify the effect of the model, we collected and produced a dataset of hyperintensity MRI liver-stage images containing over 400 cases of liver nodules. Experimental results on both public and proprietary datasets demonstrate the superiority of SUNet++ in small-scale target segmentation of three-dimensional medical images.

Feature interaction network based on hierarchical decoupled convolution for 3D medical image segmentation.

Manual image segmentation consumes time. An automatic and accurate method to segment multimodal brain tumors using context information rich three-dimensional medical images that can be used for clinical treatment decisions and surgical planning is required. However, it is a challenge to use deep learning to achieve accurate segmentation of medical images due to the diversity of tumors and the complex boundary interactions between sub-regions while limited computing resources hinder the construction of efficient neural networks. We propose a feature fusion module based on a hierarchical decoupling convolution network and an attention mechanism to improve the performance of network segmentation. We replaced the skip connections of U-shaped networks with a feature fusion module to solve the category imbalance problem, thus contributing to the segmentation of more complicated medical images. We introduced a global attention mechanism to further integrate the features learned by the encoder and explore the context information. The proposed method was evaluated for enhance tumor, whole tumor, and tumor core, achieving Dice similarity coefficient metrics of 0.775, 0.900, and 0.827, respectively, on the BraTS 2019 dataset and 0.800, 0.902, and 0.841, respectively on the BraTS 2018 dataset. The results show that our proposed method is inherently general and is a powerful tool for brain tumor image studies. Our code is available at: https://github.com/WSake/Feature-interaction-network-based-on-Hierarchical-Decoupled-Convolution.

Inflating 2D convolution weights for efficient generation of 3D medical images

Background and Objective: The generation of three-dimensional (3D) medical images has great application potential since it takes into account the 3D anatomical structure. Two problems prevent effective training of a 3D medical generative model: (1) 3D medical images are expensive to acquire and annotate, resulting in an insufficient number of training images, and (2) a large number of parameters are involved in 3D convolution. Methods: We propose a novel GAN model called 3D Split&Shuffle-GAN. To address the 3D data scarcity issue, we first pre-train a two-dimensional (2D) GAN model using abundant image slices and inflate the 2D convolution weights to improve the initialization of the 3D GAN. Novel 3D network architectures are proposed for both the generator and discriminator of the GAN model to significantly reduce the number of parameters while maintaining the quality of image generation. Several weight inflation strategies and parameter-efficient 3D architectures are investigated. Results: Experiments on both heart (Stanford AIMI Coronary Calcium) and brain (Alzheimer’s Disease Neuroimaging Initiative) datasets show that our method leads to improved 3D image generation quality (14.7 improvements on Frchet inception distance) with significantly fewer parameters (only 48.5% of the baseline method). Conclusions: We built a parameter-efficient 3D medical image generation model. Due to the efficiency and effectiveness, it has the potential to generate high-quality 3D brain and heart images for real use cases.

A Review of 3D-2D Registration Methods and Applications based on Medical Images

The registration of preoperative three-dimensional (3D) medical images with intraoperative two-dimensional (2D) data is a key technology for image-guided radiotherapy, minimally invasive surgery, and interventional procedures. In this paper, we review 3D-2D registration methods using computed tomography (CT) and magnetic resonance imaging (MRI) as preoperative 3D images and ultrasound, X-ray, and visible light images as intraoperative 2D images. The 3D-2D registration techniques are classified into intensity-based, structure-based, and gradient-based according to the different registration features. In addition, we investigated the different application scenarios of this registration technology in medical clinical treatment, which can be divided into disease diagnosis, surgical guidance and postoperative evaluation, and also investigated the evaluation method of 3D-2D registration effect.

Significantly reducing the presurgical preparation time for anterior pelvic fracture surgery by faster creating patient-specific curved plates

BackgroundTo shorten the preoperative preparation time, reconstruction plates were designed using the computed tomography (CT)-based three-dimensional (3D) medical imaging surgical planning software OOOPDS. In addition, 3D printing was used to generate curved plates for anterior pelvic fracture surgeries.MethodsThis study analyzed two groups with the same 21 patients who underwent surgery for traumatic anterior pelvic ring fractures. In Group 1, the direct reconstruction plates were preoperatively contoured according to the anatomical 3D-printed pelvic model. In Group 2, the fixation plates were contoured according to the 3D printed plate templates, which were created based on the simulated plate templates by the OOOPDS software. The processing time, including the 3D printing time for the pelvic models in Group 1, the 3D printing time for the fixation plate templates in Group 2, and the pre-contouring time for the plates in both groups, was recorded.ResultsThe mean time of pre-contouring for the curved reconstruction plates in Group 2 was significantly less than in Group 1 (−55 min; P < 0.01). The mean time of 3D printing for the 3D plate template model in Group 2 was significantly less than that for the 3D pelvic model in Group 1 (−869 min; P < 0.01). Experimental results showed that the printing time for the plate pre-contouring and the 3D plate templates could be effectively reduced by approximately 93% and 90%, respectively.ConclusionThis method can shorten the preoperative preparation time significantly.

Swin Unet3D: a three-dimensional medical image segmentation network combining vision transformer and convolution

BackgroundSemantic segmentation of brain tumors plays a critical role in clinical treatment, especially for three-dimensional (3D) magnetic resonance imaging, which is often used in clinical practice. Automatic segmentation of the 3D structure of brain tumors can quickly help physicians understand the properties of tumors, such as the shape and size, thus improving the efficiency of preoperative planning and the odds of successful surgery. In past decades, 3D convolutional neural networks (CNNs) have dominated automatic segmentation methods for 3D medical images, and these network structures have achieved good results. However, to reduce the number of neural network parameters, practitioners ensure that the size of convolutional kernels in 3D convolutional operations generally does not exceed 7 times 7 times 7, which also leads to CNNs showing limitations in learning long-distance dependent information. Vision Transformer (ViT) is very good at learning long-distance dependent information in images, but it suffers from the problems of many parameters. What’s worse, the ViT cannot learn local dependency information in the previous layers under the condition of insufficient data. However, in the image segmentation task, being able to learn this local dependency information in the previous layers makes a big impact on the performance of the model.MethodsThis paper proposes the Swin Unet3D model, which represents voxel segmentation on medical images as a sequence-to-sequence prediction. The feature extraction sub-module in the model is designed as a parallel structure of Convolution and ViT so that all layers of the model are able to adequately learn both global and local dependency information in the image.ResultsOn the validation dataset of Brats2021, our proposed model achieves dice coefficients of 0.840, 0.874, and 0.911 on the ET channel, TC channel, and WT channel, respectively. On the validation dataset of Brats2018, our model achieves dice coefficients of 0.716, 0.761, and 0.874 on the corresponding channels, respectively.ConclusionWe propose a new segmentation model that combines the advantages of Vision Transformer and Convolution and achieves a better balance between the number of model parameters and segmentation accuracy. The code can be found at https://github.com/1152545264/SwinUnet3D.