Segmentation Of Skeleton Research Articles

Skeleton-guided 3D convolutional neural network for tubular structure segmentation.

Accurate segmentation of tubular structures is crucial for clinical diagnosis and treatment but is challenging due to their complex branching structures and volume imbalance. The purpose of this study is to propose a 3D deep learning network that incorporates skeleton information to enhance segmentation accuracy in these tubular structures. Our approach employs a 3D convolutional network to extract 3D tubular structures from medical images such as CT volumetric images. We introduce a skeleton-guided module that operates on extracted features to capture and preserve the skeleton information in the segmentation results. Additionally, to effectively train our deep model in leveraging skeleton information, we propose a sigmoid-adaptive Tversky loss function which is specifically designed for skeleton segmentation. We conducted experiments on two distinct 3D medical image datasets. The first dataset consisted of 90 cases of chest CT volumetric images, while the second dataset comprised 35 cases of abdominal CT volumetric images. Comparative analysis with previous segmentation approaches demonstrated the superior performance of our method. For the airway segmentation task, our method achieved an average tree length rate of 93.0%, a branch detection rate of 91.5%, and a precision rate of 90.0%. In the case of abdominal artery segmentation, our method attained an average precision rate of 97.7%, a recall rate of 91.7%, and an F-measure of 94.6%. We present a skeleton-guided 3D convolutional network to segment tubular structures from 3D medical images. Our skeleton-guided 3D convolutional network could effectively segment small tubular structures, outperforming previous methods.

Automatic Skeleton Segmentation in CT Images Based on U-Net

Bone metastasis, emerging oncological therapies, and osteoporosis represent some of the distinct clinical contexts which can result in morphological alterations in bone structure. The visual assessment of these changes through anatomical images is considered suboptimal, emphasizing the importance of precise skeletal segmentation as a valuable aid for its evaluation. In the present study, a neural network model for automatic skeleton segmentation from bidimensional computerized tomography (CT) slices is proposed. A total of 77 CT images and their semimanual skeleton segmentation from two acquisition protocols (whole-body and femur-to-head) are used to form a training group and a testing group. Preprocessing of the images includes four main steps: stretcher removal, thresholding, image clipping, and normalization (with two different techniques: interpatient and intrapatient). Subsequently, five different sets are created and arranged in a randomized order for the training phase. A neural network model based on U-Net architecture is implemented with different values of the number of channels in each feature map and number of epochs. The model with the best performance obtains a Jaccard index (IoU) of 0.959 and a Dice index of 0.979. The resultant model demonstrates the potential of deep learning applied in medical images and proving its utility in bone segmentation.

Leveraging Model Scaling and Butterfly Network in the Bone Scan Image Segmentation

As we all know, cancer is one of the leading causes of death worldwide and the second leading cause of death overall. This is why regular screenings or health checks are necessary to detect cancer lesions early. Since bone scan images have become the primary means of detecting the emergence of cancer lesions on bone, high segmentation accuracy is essential for establishing the model of some predefined regions in bone scan images where cancer metastasis was predicted to appear. Consequently, robust localization and identification of the specific region in bone scan images are required for automated metastasis detection. To this end, we propose Efficient-BtrflyNet, a new deep learning-based architecture for skeleton segmentation of whole-body bone scan images. The proposed architecture exploits the benefits of EfficientNet’s model scaling and the encoder–decoder design of butterfly-type networks. We added EfficientNetB7 to the encoder section to obtain more specific features. The proposed architecture simultaneously processes anterior and posterior whole-body bone scan images. Using 37 bone scan images, we evaluated the performance of our proposed skeleton segmentation system using the Dice score. Efficient-BtrflyNet achieves superior segmentation performance compared to the existing representative method.

Deep learning assisted atlas-based delineation of the skeleton from Whole-Body Diffusion Weighted MRI in patients with malignant bone disease

BackgroundWhole-Body Diffusion Weighted MRI (WBDWI) can provide a basis for quantifying disease volume and apparent diffusion coefficient (ADC) to aid disease staging and response assessment of metastatic bone disease. In this study, we develop atlas-based registration assisted by two AI-based models (i) body-region classification for automatic identification of legs, pelvis, lumbar/thoracic/cervical spine and head, and (ii) automated spinal cord segmentation, to automatically delineate the whole skeleton on WBDWI. MethodsTraining and validation of AI-based models were performed on 40 patients (32:8 split) with confirmed Advanced Prostate Cancer (APC) who underwent baseline and after treatment WBDWI scans. Consequent atlas-based skeleton segmentation accuracy was validated using leave-one-out cross-validation in a cohort of 15 patients with Multiple Myeloma (MM). ResultsAtlas-based registration in combination with AI-based models demonstrated higher performance in segmenting the spinal column when compared with the traditional techniques (average dice score for the lumbar, thoracic and cervical spine of 0.77/0.78/0.68 compared to 0.71/0.69/0.51 from the conventional method, respectively). Similar performances were observed for long bones, pelvic girdle and ribcage (average dice score of 0.65/0.7/0.47, respectively). Our approach demonstrated lower relative median ADC and skeleton volume difference between estimation and ground truth compared to conventional atlas-based methods (average relative median ADC and skeleton volume difference of 10.3/2.1% compared to 13/2.7%, respectively). ConclusionAI-augmented atlas-based registration substantially improves the accuracy of derived skeletal segmentations from WBDWI compared with traditional atlas-based registration.

Vision based crop row navigation under varying field conditions in arable fields

Accurate crop row detection is often challenged by the varying field conditions present in real-world arable fields. Traditional colour based segmentation is unable to cater for all such variations. The lack of comprehensive datasets in agricultural environments limits the researchers from developing robust segmentation models to detect crop rows. We present a dataset for crop row detection with 11 field variations from sugar beet and maize crops. We also present a novel crop row detection algorithm for visual servoing in crop row fields. The proposed method uses deep learning based crop row skeleton segmentation method followed by a crop row scanning algorithm that identifies the central crop row which the robot then follows. The unique dataset we used with skeleton representations for crop row detection enables robust crop row detection in challenging real world field conditions. Our algorithm can detect crop rows against varying field conditions such as curved crop rows, weed presence, discontinuities, growth stages, tramlines, shadows and light levels. Dense weed presence within inter-row space and discontinuities in crop rows were the most challenging field conditions for our crop row detection algorithm. An End-of Row detector algorithm was developed to detect the end of the crop row and navigate the robot towards the headland area when it reaches the end of the crop row.

Leveraging topology for domain adaptive road segmentation in satellite and aerial imagery

Getting precise aspects of road through segmentation from remote sensing imagery is useful for many real-world applications such as autonomous vehicles, urban development and planning, and achieving sustainable development goals (SDGs).11https://sdgs.un.org/goals. Roads are only a small part of the image, and their appearance, type, width, elevation, directions, etc. exhibit large variations across geographical areas. Furthermore, due to differences in urbanization styles, planning, and the natural environments; regions along the roads vary significantly. Due to these variations among the train and test domains (domain shift), the road segmentation algorithms fail to generalize to new geographical locations. Unlike the generic domain alignment scenarios, road segmentation has no scene structure and generic domain adaptive segmentation methods are unable to enforce topological properties like continuity, connectivity, smoothness, etc., thus resulting in degraded domain alignment. In this work, we propose a topology-aware unsupervised domain adaptation approach for road segmentation in remote sensing imagery. During domain adaptation for road segmentation, we predict road skeleton, an auxiliary task to enforce the topological constraints. To enforce consistent predictions of road and skeleton, especially in the unlabeled target domain, the conformity loss is defined across the skeleton prediction head and the road-segmentation head. Furthermore, for self-training, we filter out the noisy pseudo-labels by using a connectivity-based pseudo-labels refinement strategy, on both road and skeleton segmentation heads, thus avoiding holes and discontinuities. Extensive experiments on the benchmark datasets show the effectiveness of the proposed approach compared to existing state-of-the-art methods. Specifically, for SpaceNet to DeepGlobe adaptation, the proposed approach outperforms the competing methods by a minimum margin of 6.6%, 6.7%, and 9.8% in IoU, F1-score, and APLS, respectively. (The source code is available on Github).

Bone Metastases Lesion Segmentation on Breast Cancer Bone Scan Images with Negative Sample Training.

The use of deep learning methods for the automatic detection and quantification of bone metastases in bone scan images holds significant clinical value. A fast and accurate automated system for segmenting bone metastatic lesions can assist clinical physicians in diagnosis. In this study, a small internal dataset comprising 100 breast cancer patients (90 cases of bone metastasis and 10 cases of non-metastasis) and 100 prostate cancer patients (50 cases of bone metastasis and 50 cases of non-metastasis) was used for model training. Initially, all image labels were binary. We used the Otsu thresholding method or negative mining to generate a non-metastasis mask, thereby transforming the image labels into three classes. We adopted the Double U-Net as the baseline model and made modifications to its output activation function. We changed the activation function to SoftMax to accommodate multi-class segmentation. Several methods were used to enhance model performance, including background pre-processing to remove background information, adding negative samples to improve model precision, and using transfer learning to leverage shared features between two datasets, which enhances the model's performance. The performance was investigated via 10-fold cross-validation and computed on a pixel-level scale. The best model we achieved had a precision of 69.96%, a sensitivity of 63.55%, and an F1-score of 66.60%. Compared to the baseline model, this represents an 8.40% improvement in precision, a 0.56% improvement in sensitivity, and a 4.33% improvement in the F1-score. The developed system has the potential to provide pre-diagnostic reports for physicians in final decisions and the calculation of the bone scan index (BSI) with the combination with bone skeleton segmentation.

Background separation of sports athletes and motion image analysis based on skeleton segmentation algorithm

Background separation of sports athletes and motion image analysis based on skeleton segmentation algorithm

Topological-Preserving Membrane Skeleton Segmentation in Multiplex Immunofluorescence Imaging.

Multiplex immunofluorescence (MxIF) is an emerging imaging technology whose downstream molecular analytics highly rely upon the effectiveness of cell segmentation. In practice, multiple membrane markers (e.g., NaKATPase, PanCK and β-catenin) are employed to stain membranes for different cell types, so as to achieve a more comprehensive cell segmentation since no single marker fits all cell types. However, prevalent watershed-based image processing might yield inferior capability for modeling complicated relationships between markers. For example, some markers can be misleading due to questionable stain quality. In this paper, we propose a deep learning based membrane segmentation method to aggregate complementary information that is uniquely provided by large scale MxIF markers. We aim to segment tubular membrane structure in MxIF data using global (membrane markers z-stack projection image) and local (separate individual markers) information to maximize topology preservation with deep learning. Specifically, we investigate the feasibility of four SOTA 2D deep networks and four volumetric-based loss functions. We conducted a comprehensive ablation study to assess the sensitivity of the proposed method with various combinations of input channels. Beyond using adjusted rand index (ARI) as the evaluation metric, which was inspired by the clDice, we propose a novel volumetric metric that is specific for skeletal structure, denoted as . In total, 80 membrane MxIF images were manually traced for 5-fold cross-validation. Our model outperforms the baseline with a 20.2% and 41.3% increase in and ARI performance, which is significant (p<0.05) using the Wilcoxon signed rank test. Our work explores a promising direction for advancing MxIF imaging cell segmentation with deep learning membrane segmentation. Tools are available at https://github.com/MASILab/MxIF_Membrane_Segmentation.

An end-to-end multi-task system of automatic lesion detection and anatomical localization in whole-body bone scintigraphy by deep learning.

Limited by spatial resolution and visual contrast, bone scintigraphy interpretation is susceptible to subjective factors, which considerably affects the accuracy and repeatability of lesion detection and anatomical localization. In this work, we design and implement an end-to-end multi-task deep learning model to perform automatic lesion detection and anatomical localization in whole-body bone scintigraphy. A total of 617 whole-body bone scintigraphy cases including anterior and posterior views were retrospectively analyzed. The proposed semi-supervised model consists of two task flows. The first one, the lesion segmentation flow, received image patches and was trained in a supervised way. The other one, skeleton segmentation flow, was trained on as few as five labeled images in conjunction with the multi-atlas approach, in a semi-supervised way. The two flows joint in their encoder layers so each flow can capture more generalized distribution of the sample space and extract more abstract deep features. The experimental results show that the architecture achieved the highest precision in the finest bone segmentation task in both anterior and posterior images of whole-body scintigraphy. Such an end-to-end approach with very few manual annotation requirement would be suitable for algorithm deployment. Moreover, the proposed approach reliably balances unsupervised labels construction and supervised learning, providing useful insight for weakly labeled image analysis. Supplementary data are available at Bioinformatics online.

An automatic fine-grained skeleton segmentation method for whole-body bone scintigraphy using atlas-based registration.

Whole-body bone scintigraphy (WBS) is one of the common imaging methods in nuclear medicine. It is a time-consuming, tedious, and error-prone issue for physicians to determine the location of bone lesions which is important for the qualitative diagnosis of bone lesions. In this paper, an automatic fine-grained skeleton segmentation method for WBS is developed. The proposed method contains four steps. In the first step, a novel denoising method is proposed to remove the noise from WBS which benefits the location of the skeleton. In the second step, a restoration method based on gray probability distribution is developed to repair the partial contamination caused by the high local density of radionuclide. Then, the standardization for WBS is performed by the exact histogram matching. Finally, the deformation field between the atlas and the input WBS is calculated by registration, and the segmentation mask of the input WBS is obtained by wrapping the segmentation mask of the atlas with the deformation field. The experimental results show that the proposed method outperforms the traditional registration (Morphon): mean square error decreased from [Formula: see text] to [Formula: see text], peak signal-to-noise ratio increased from 21.26 to 26.92, and mean structural similarity increased from 0.9986 to 0.9998. Our experiments show that the proposed method can achieve robust and fine-grained results which outperform the traditional registration method, indicating it could be helpful in clinical application. To the best of our knowledge, this is the first work that implements a fully automated fine-grained skeleton segmentation method for WBS.

Correlation between whole skeleton dual energy CT calcium-subtracted attenuation and bone marrow infiltration in multiple myeloma

ObjectivesObjective evaluation of the extent of skeletal marrow involvement in multiple myeloma remains a clinical gap for CT. We aimed to develop a quantitative segmentation pipeline for dual energy CT and to assess whether quantified whole skeleton calcium-subtracted attenuation values correlate with biopsy-derived bone marrow infiltration in multiple myeloma. MethodsConsecutive prospective patients with suspected/established myeloma underwent dual source CT from the skull vertex to proximal tibia. Whole skeleton segmentation was performed for 120 kVp-equivalent images as follows: following Hounsfield unit (HU) thresholding, a Chan-Vese morphological operation was implemented to generate a whole skeleton segmentation mask. This mask was then applied to corresponding whole skeleton material decomposition calcium-subtracted maps, generating whole skeleton HU values. Associations with biopsy-derived bone marrow plasma cell infiltration percentage were assessed with Spearman’s rank correlation; significance was at 5%. Results21 patients (12 females; median (IQR) 67 (61, 73) years) were included; 16 patients had osteolytic bone lesions; 15 patients underwent bone marrow biopsy. Segmentation and quantification were feasible in all patients. Median (IQR) of the average skeletal calcium-subtracted attenuation was −59.9 HU (-66.3, −51.8HU). There was a positive correlation with bone marrow plasma cell infiltration percentage (Spearman’s rho: + 0.79, p < 0.001). ConclusionWhole skeleton calcium-subtracted attenuation is associated with the degree of bone marrow infiltration by plasma cells, providing an objective measure of marrow involvement with the potential to allow earlier detection of disease.

INTERIOR DECORATION OF AUGUSTE PERRET'S "BOTTLE" AND LE CORBUSIER'S "EQUIPMENT" :

This paper aims to clarify cross points of interior decoration in Auguste Perret (1874-1954) and Le Corbusier (Charles-Edouard Jeanneret, 1887-1965) as the considerations concerning the concept of "decor" in the 20th century. As a result of chronological reorganization of the training of Jeanneret at the Perret’s office and the subsequent communication between two mainly by correspendances, the difference between the two became clear in the discussion on "movable furniture" and "immovable furniture". It was also a question of strategy regarding the segmentation of structural skeleton and furniture.

A comprehensive series of Irx cluster mutants reveals diverse roles in facial cartilage development.

Proper function of the vertebrate skeleton requires the development of distinct articulating embryonic cartilages. Irx transcription factors are arranged in co-regulated clusters that are expressed in the developing skeletons of the face and appendages. IrxB cluster genes are required for the separation of toes in mice and formation of the hyoid joint in zebrafish, yet whether Irx genes have broader roles in skeletal development remains unclear. Here, we perform a comprehensive loss-of-function analysis of all 11 Irx genes in zebrafish. We uncover conserved requirements for IrxB genes in formation of the fish and mouse scapula. In the face, we find a requirement for IrxAb genes and irx7 in formation of anterior neural crest precursors of the jaw, and for IrxBa genes in formation of endodermal pouches and gill cartilages. We also observe extensive joint loss and cartilage fusions in animals with combinatorial losses of Irx clusters, with in vivo imaging revealing that at least some of these fusions arise through inappropriate chondrogenesis. Our analysis reveals diverse roles for Irx genes in the formation and later segmentation of the facial skeleton.

Concrete Surface Crack Recognition in Complex Scenario Based on Deep Learning

A concrete crack detection method suitable for various scene conditions is proposed on the basis of image recognition to complete the classification and the identification of the cracks of the collected crack images in the bridge health monitoring work conveniently and reliably. Moreover, this method can improve the crack recognition effect, which is greatly affected by the selection of the initial clustering center of the extraction algorithm, and has high environmental dependence on the image background. Convolutional neural networks (CNNs) is a representative deep learning algorithm that can characterize learning and classify input information according to its own hierarchical structure. After collecting image data on the spot by relying on the Baoxie River Bridge inspection project, Gaoxin 4th Road, Donghu High-tech Zone, Wuhan City based on the CNN, an image classification model suitable for concrete crack image classification is established. This model realizes the collection of the inspection project image classification of concrete structures in complex scenes while considering that the local cluster density and Euclidean distance of the clustering center in the traditional K-means algorithm are both large. The traditional K-means algorithm is improved by combining the use of statistical principles and morphological methods. Finally, the improved K-means algorithm completes the crack skeleton segmentation extraction and crack width calculation of crack images in complex scenes. The effectiveness of the proposed method under concrete surface peelings, stains, mosses, or other noise conditions was verified according to the successful identification of 600 crack on-site images photographed from a bridge surface. Results also show that the efficiency of the proposed crack detection method is higher than that of traditional methods. The proposed approach can also provide a reference for the in depth research on crack identification on the surface of concrete structures in complex background in the future.

Automated 3D MRI rendering of the craniofacial skeleton: using ZTE to drive the segmentation of black bone and FIESTA-C images

PurposeAutomated bone segmentation from MRI datasets would have a profound impact on clinical utility, particularly in the craniofacial skeleton where complex anatomy is coupled with radiosensitive organs. Techniques such as gradient echo black bone (GRE-BB) and short echo time (UTE, ZTE) have shown potential in this quest. The objectives of this study were to ascertain (1) whether the high-contrast of zero echo time (ZTE) could drive segmentation of high-resolution GRE-BB data to enhance 3D-output and (2) if these techniques could be extrapolated to ZTE driven segmentation of a routinely used non bone-specific sequence (FIESTA-C).MethodsEleven adult volunteers underwent 3T MRI examination with sequential acquisition of ZTE, GRE-BB and FIESTA-C imaging. Craniofacial bone segmentation was performed using a fully automated segmentation algorithm. Segmentation was completed individually for GRE-BB and a modified version of the algorithm was subsequently implemented, wherein the bone mask yielded by ZTE segmentation was used to initialise segmentation of GRE-BB. The techniques were subsequently applied to FIESTA-C datasets. The resulting 3D reconstructions were evaluated for areas of unexpected bony defects and discrepancies.ResultsThe automated segmentation algorithm yielded acceptable 3D outputs for all GRE-BB datasets. These were enhanced with the modified algorithm using ZTE as a driver, with improvements in areas of air/bone interface and dense muscular attachments. Comparable results were obtained with ZTE+FIESTA-C.ConclusionAutomated 3D segmentation of the craniofacial skeleton is enhanced through the incorporation of a modified segmentation algorithm utilising ZTE. These techniques are transferrable to FIESTA-C imaging which offers reduced acquisition time and therefore improved clinical utility.

Automated diagnosis of bone metastasis based on multi-view bone scans using attention-augmented deep neural networks.

Bone scintigraphy is accepted as an effective diagnostic tool for whole-body examination of bone metastasis. However, the manual analysis of bone scintigraphy images requires extensive experience and is exhausting and time-consuming. An automated diagnosis system for such images is therefore much desired. Although automatic or semi-automatic methods for the diagnosis of bone scintigraphy images have been widely studied, they employ various steps to classify the images, including segmentation of the entire skeleton, detection of hot spots, and feature extraction, which are complex and inadequately validated on small datasets, thereby resulting in low accuracy and reliability. In this paper, we describe the development of a deep convolutional neural network to determine the absence or presence of bone metastasis. This model consisting of three sub-networks that aim to extract, aggregate, and classify high-level features in a data-driven manner. There are two main innovations behind this method; First, the diagnosis is performed by jointly analyzing both anterior and posterior views, which leads to high accuracy. Second, a spatial attention feature aggregation operator is proposed to enhance the spatial location information. A large annotated bone scintigraphy image dataset containing 15,474 examinations from 13,811 patients was constructed to train and evaluate the model. The proposed method is compared with three human experts. The high classification accuracy achieved demonstrates the effectiveness of the proposed architecture for the diagnosis of bone scintigraphy images, and that it can be applied as a clinical decision support tool.

A new technique for the 3D reconstruction of the incisional hernia: A pilot study

PurposeThis study aimed to accurately evaluate incisional hernias with a new three-dimensional (3D) reconstruction technology, proving surgeons more information for intuitive and accurate judgments about incisional hernia to minimize the perioperative complications and recurrence rate. MethodThis was a pilot study using a new technique, 3D reconstruction, based on computed tomography (CT) scans to measure abdominal wall defect, herniary area, herniary volume, abdominal cavity volume, and the volume of transverse, oblique, and recti abdominis in three patients with incisional hernias. ResultsThe 3D reconstruction technique made automated segmentation of the bony skeleton, skin, outer abdominal wall, vessel, and hernia sac. The hernia sac, abdominal muscles, and their anatomic relationship were clearly illustrated in 3D reconstruction images. Moreover, abdominal cavity volume; herniary diameter, area, and volume; and the volume of transverse, oblique, and recti muscles could be evaluated through 3D reconstruction images. Surgeons can also freely combine, rotate, scale, and move the 3D reconstruction mode, modify the name and transparency of the 3D reconstruction model, and observe the internal structure of the tissue and the size, shape, and location of the lesion from multiple angles for better and accurate judgments. ConclusionThe herniary diameter, area, and volume and the volume of transverse, oblique, and recti abdominis can be accurately calculated through this 3D reconstruction technology. A three-dimensional vision of the abdomen through this technology can objectively and quantitatively evaluate the situation of incisional hernia, providing a more realistic means for diagnosis and treatment of incisional hernias.

Dstyk mutation leads to congenital scoliosis-like vertebral malformations in zebrafish via dysregulated mTORC1/TFEB pathway

Congenital scoliosis (CS) is a complex genetic disorder characterized by vertebral malformations. The precise etiology of CS is not fully defined. Here, we identify that mutation in dual serine/threonine and tyrosine protein kinase (dstyk) lead to CS-like vertebral malformations in zebrafish. We demonstrate that the scoliosis in dstyk mutants is related to the wavy and malformed notochord sheath formation and abnormal axial skeleton segmentation due to dysregulated biogenesis of notochord vacuoles and notochord function. Further studies show that DSTYK is located in late endosomal/lysosomal compartments and is involved in the lysosome biogenesis in mammalian cells. Dstyk knockdown inhibits notochord vacuole and lysosome biogenesis through mTORC1-dependent repression of TFEB nuclear translocation. Inhibition of mTORC1 activity can rescue the defect in notochord vacuole biogenesis and scoliosis in dstyk mutants. Together, our findings reveal a key role of DSTYK in notochord vacuole biogenesis, notochord morphogenesis and spine development through mTORC1/TFEB pathway.

Semantic Segmentation of Eight Regions of Upper and Lower Limb Bones Using 3D U-Net in Whole-body CT Images

Automated analysis of skeletal muscle in whole-body computed tomography (CT) images uses bone information, but bone segmentation including the epiphysis is not achieved. The purpose of this research was the semantic segmentation of eight regions of upper and lower limb bones including the epiphysis in whole-body CT images. Our targets were left and right upper arms, forearms, thighs, and lower legs. We connected two 3D U-Nets in cascade for segmentation of eight upper and lower limb bones in whole-body CT images. The first 3D U-Net was used for skeleton segmentation in whole-body CT images, and the second 3D U-Net was used for eight upper and lower limb bones' segmentation in skeleton segmentation results. Thirty cases of whole-body CT images were used in the experiment, and the segmentation results were evaluated using Dice coefficient with 3-fold cross-validation. The mean Dice coefficient was 93% in the left and right upper arms, 89% in the left and right forearms, 95% in the left and right thighs, and 94% in the left and right lower legs. Although the accuracy of the segmentation results of relatively small bones remains a challenge, the semantic segmentation of eight regions of upper and lower limb bones including the epiphysis in whole-body CT images has been achieved.