Volumetric Overlap Error Research Articles

Fully Automated Detection of the Appendix Using U-Net Deep Learning Architecture in CT Scans

Background: The accurate segmentation of the appendix with well-defined boundaries is critical for diagnosing conditions such as acute appendicitis. The manual identification of the appendix is time-consuming and highly dependent on the expertise of the radiologist. Method: In this study, we propose a fully automated approach to the detection of the appendix using deep learning architecture based on the U-Net with specific training parameters in CT scans. The proposed U-Net architecture is trained on an annotated original dataset of abdominal CT scans to segment the appendix efficiently and with high performance. In addition, to extend the training set, data augmentation techniques are applied for the created dataset. Results: In experimental studies, the proposed U-Net model is implemented using hyperparameter optimization and the performance of the model is evaluated using key metrics to measure diagnostic reliability. The trained U-Net model achieved the segmentation performance for the detection of the appendix in CT slices with a Dice Similarity Coefficient (DSC), Volumetric Overlap Error (VOE), Average Symmetric Surface Distance (ASSD), Hausdorff Distance 95 (HD95), Precision (PRE) and Recall (REC) of 85.94%, 23.29%, 1.24 mm, 5.43 mm, 86.83% and 86.62%, respectively. Moreover, our model outperforms other methods by leveraging the U-Net’s ability to capture spatial context through encoder–decoder structures and skip connections, providing a correct segmentation output. Conclusions: The proposed U-Net model showed reliable performance in segmenting the appendix region, with some limitations in cases where the appendix was close to other structures. These improvements highlight the potential of deep learning to significantly improve clinical outcomes in appendix detection.

Computer-aided diagnosis of cystic lung diseases using CT scans and deep learning.

Auxiliary diagnosis of different types of cystic lung diseases (CLDs) is important in the clinic and is instrumental in facilitating early and specific treatments. Current clinical methods heavily depend on accumulated experience, restricting their applicability in regions with less developed medical resources. Thus, how to realize the computer-aided diagnosis of CLDs is of great clinical value. This work proposed a deep learning-based method for automatically segmenting the lung parenchyma in computed tomography (CT) slice images and accurately diagnosing the CLDs using CT scans. A two-stage deep learning method was proposed for the automatic classification of normal cases and five different CLDs using CT scans. Lung parenchyma segmentation is the foundation of CT image analysis and auxiliary diagnosis. To meet the requirements of different sizes of the lung parenchyma, an adaptive region-growing and improved U-Net model was employed for mask acquisition and automatic segmentation. The former was achieved by a self-designed adaptive seed point selection method based on similarity measurement, and the latter introduced multiscale input and multichannel output into the original U-Net model and effectively achieved the lightweight design by adjusting the structure and parameters. After that, the middle 30 consecutive CT slice images of each sample were segmented to obtain lung parenchyma, which was employed for training and testing the proposed multichannel parallel input recursive MLP-Mixer network (MPIRMNet) model, achieving the computer-aided diagnosis of CLDs. A total of 4718 and 16290 CT slice images collected from 543 patients were employed to validate the proposed segmentation and classification methods, respectively. Experimental results showed that the improved U-Net model can accurately segment the lung parenchyma in CT slice images, with the Dice, precision, volumetric overlap error, and relative volume difference of 0.96±0.01, 0.93±0.04, 0.05±0.02, and 0.05±0.03, respectively. Meanwhile, the proposed MPIRMNet model achieved appreciable classification effect for normal cases and different CLDs, with the accuracy, sensitivity, specificity, and F1 score of 0.8823±0.0324, 0.8897±0.0325, 0.9746±0.0078, and 0.8831±0.0334, respectively. Compared with classical machine learning and convolutional neural networks-based methods for this task, the proposed classification method had a preferable performance, with a significant improvement of accuracy of 10.74%. The work introduced a two-stage deep learning method, which can achieve the segmentation of lung parenchyma and the classification of CLDs. Compared to previous diagnostic tasks targeting single CLD, this work can achieve various CLDs' diagnosis in the early stage, thereby achieving targeted treatment and increasing the potential and value of clinical applications.

Construction of multi-scale feature fusion segmentation model of MRI knee images based on dual attention mechanism weighted aggregation.

Early diagnosis of knee osteoarthritis is an important area of research in the field of clinical medicine. Due to the complexity in the MRI imaging sequences and the diverse structure of cartilage, there are many challenges in the segmentation of knee bone and cartilage. Relevant studies have conducted semantic fusion processing through splicing or summing forms, which results in reduced resolution and the accumulation of redundant information. This study was envisaged to construct an MRI image segmentation model to improve the diagnostic efficiency and accuracy of different grade knee osteoarthritis by adopting the Dual Attention and Multi-scale Feature Fusion Segmentation network (DA-MFFSnet). The feature information of different scales was fused through the Multi-scale Attention Downsample module to extract more accurate feature information, and the Global Attention Upsample module weighted lower-level feature information to reduce the loss of key information. The collected MRI knee images were screened and labeled, and the study results showed that the segmentation effect of DA-MFFSNet model was closer to that of the manually labeled images. The mean intersection over union, the dice similarity coefficient and the volumetric overlap error was 92.74%, 91.08% and 7.44%, respectively, and the accuracy of the differential diagnosis of knee osteoarthritis was 84.42%. The model exhibited better stability and classification effect. Our results indicated that the Dual Attention and Multi-scale Feature Fusion Segmentation model can improve the segmentation effect of MRI knee images in mild and medium knee osteoarthritis, thereby offering an important clinical value and improving the accuracy of the clinical diagnosis.

Deep Learning Framework for Liver Tumor Segmentation

INTRODUCTION: Segregating hepatic tumors from the liver in computed tomography (CT) scans is vital in hepatic surgery planning. Extracting liver tumors in CT images is complex due to the low contrast between the malignant and healthy tissues and the hazy boundaries in CT images. Moreover, manually detecting hepatic tumors from CT images is complicated, time-consuming, and needs clinical expertise.
 OBJECTIVES: An automated liver and hepatic malignancies segmentation is essential to improve surgery planning, therapy, and follow-up evaluation. Therefore, this study demonstrates the creation of an intuitive approach for segmenting tumors from the liver in CT scans.
 METHODS: The proposed framework uses residual UNet (ResUNet) architecture and local region-based segmentation. The algorithm begins by segmenting the liver, followed by malignancies within the liver envelope. First, ResUNet trained on labeled CT images predicts the coarse liver pixels. Further, the region-level segmentation helps determine the tumor and improves the overall segmentation map. The model is tested on a public 3D-IRCADb dataset.
 RESULTS: Two metrics, namely dice coefficient and volumetric overlap error (VOE), were used to evaluate the performance of the proposed method. ResUNet model achieved dice of 0.97 and 0.96 in segmenting liver and tumor, respectively. The value of VOE is also reduced to 1.90 and 0.615 for liver and tumor segmentation.
 CONCLUSION: The proposed ResUNet model performs better than existing methods in the literature. Since the proposed model is built using U-Net, the model ensures quality and precise dimensions of the output.

Sparse deep belief network coupled with extended local fuzzy active contour model-based liver cancer segmentation from abdomen CT images.

Liver cancer from abdominal CT images must be accurately segmented for the purpose of diagnosis with treatment planning. But, the similarity in gray values between the liver and the surrounding tissues poses a challenge. To address this, a novel sparse deep belief network coupled with extended local fuzzy active contour model-based liver cancer segmentation from abdomen CT images (SDBN-ELFAC-LCS-CT) is proposed. This method incorporates dynamic adaptive pooling and residual modules in SDBN to improve the feature selection and generalization ability. Additionally, the 3D reconstruction is performed to refine segmentation results. The proposed SDBN-ELFAC-LCS-CT approach is implemented in MATLAB. The performance of the proposed SDBN-ELFAC-LCS-CT achieves dice coefficients that were up to 96.16% higher and 75.88%, 88.75%, and 71.16% lower. Volumetric overlap error compared with existing models, like basic ensembles of vanilla-style deep learning modes, increases liver segmentation from CT imageries (BEVS-LCS-CT), an incorporated 3 dimensional sparse deep belief network along enriched seagull optimization approach for liver segmentation (3DBN-ESOA-LCS-CT) and iterative convolutional encoder-decoder network and multiple scale context learning for segmenting liver (ICEDN-LCS-CT), respectively.

CotepRes-Net: An efficient U-Net based deep learning method of liver segmentation from Computed Tomography images

Automatic liver segmentation from CT images is challenging due to the indistinct boundaries between the liver and surrounding organs in the abdominal cavity CT. To address these limitations, we propose a deep learning model called CotepRes-Net, which is based on the U-Net architecture but with modified skip connection using a convolutional block attention module (CBAM). We introduce a modified inception block and a contextual transformer block (CoT) at the bottleneck to capture richer feature representations and leverage global contextual information to constrain the liver shape at the unclear boundary. To accelerate the network convergence in the training stage, we utilize an Efficient channel attention module (ECA) in the decoder layers and replace the convolution operation in each convolution layer with a residual structure. We evaluate the performance of the proposed model using the publicly available dataset LiTS (MICCAI 2017 Liver Tumor Segmentation Challenge) and show the segmentation results on the dataset Sliver07 (Segmentation of the Liver Competition 2007). The results show that our model achieves a dice similarity coefficient (DSC) of 0.9688, Jaccard index of 0.9422, volumetric overlap error (VOE) of 0.0578, relative absolute volume difference (RAVD) of 0.0039, average symmetric surface distance (ASSD) of 1.094 mm and maximum symmetric surface distance (MSSD) of 16.079 mm for liver segmentation. The experimental results show that the proposed method is comparable to or even outperform the state-of-the-art methods in terms of accuracy and computation cost, which can serve as a valuable tool for image-based liver analysis.

Automatic Kidney Segmentation Method Based on an Enhanced Generative Adversarial Network

When deciding on a kidney tumor's diagnosis and treatment, it is critical to take its morphometry into account. It is challenging to undertake a quantitative analysis of the association between kidney tumor morphology and clinical outcomes due to a paucity of data and the need for the time-consuming manual measurement of imaging variables. To address this issue, an autonomous kidney segmentation technique, namely SegTGAN, is proposed in this paper, which is based on a conventional generative adversarial network model. Its core framework includes a discriminator network with multi-scale feature extraction and a fully convolutional generator network made up of densely linked blocks. For qualitative and quantitative comparisons with the SegTGAN technique, the widely used and related medical image segmentation networks U-Net, FCN, and SegAN are used. The experimental results show that the Dice similarity coefficient (DSC), volumetric overlap error (VOE), accuracy (ACC), and average surface distance (ASD) of SegTGAN on the Kits19 dataset reach 92.28%, 16.17%, 97.28%, and 0.61 mm, respectively. SegTGAN outscores all the other neural networks, which indicates that our proposed model has the potential to improve the accuracy of CT-based kidney segmentation.

Automatic segmentation of ultrasound images of carotid atherosclerotic plaque based on Dense-UNet.

Carotid atherosclerosis plaque rupture is an important cause of myocardial infarction and stroke. The effective segmentation of ultrasound images of carotid atherosclerotic plaques aids clinicians to accurately assess plaque stability. At present, this procedure relies mainly on the experience of the medical practitioner to manually segment the ultrasound image of the carotid atherosclerotic plaque. This method is also time-consuming. This study intends to establish an automatic intelligent segmentation method of ultrasound images of carotid plaque. The present study combined the U-Net and DenseNet networks, to automatically segment the ultrasound images of carotid atherosclerotic plaques. The same test set was selected and segmented using the traditional U-Net network and the ResUNet network. The prediction results of the three network models were compared using Dice (Dice similarity coefficient), and VOE (volumetric overlap error) coefficients. Compared with the existing U-Net network and ResUNet network, the Dense-UNet network exhibited an optimal effect on the automated segmentation of the ultrasound images. The Dense-UNet network could realize the automatic segmentation of atherosclerotic plaque ultrasound images, and it could assist medical practitioners in plaque evaluation.

Teeth U-Net: A segmentation model of dental panoramic X-ray images for context semantics and contrast enhancement

Background and objectiveIt is very significant in orthodontics and restorative dentistry that the teeth are segmented from dental panoramic X-ray images. Nevertheless, there are some problems in panoramic X-ray images of teeth, such as blurred interdental boundaries, low contrast between teeth and alveolar bone. MethodsIn this paper, The Teeth U-Net model is proposed in this paper to resolve these problems. This paper makes the following contributions: Firstly, a Squeeze-Excitation Module is utilized in the encoder and the decoder. And proposing a dense skip connection between encoder and decoder to reduce the semantic gap. Secondly, due to the irregular shape of the teeth and the low contrast of the dental panoramic X-ray images. A Multi-scale Aggregation attention Block (MAB) in the bottleneck layer is designed to resolve this problem, which can effectively extract teeth shape features and fuse multi-scale features adaptively. Thirdly, in order to capture dental feature information in a larger field of perception, this paper designs a Dilated Hybrid self-Attentive Block (DHAB) at the bottleneck layer. This module effectively suppresses the task-irrelevant background region information without increasing the network parameters. Finally, the effectiveness of the algorithm is validated using a clinical dental panoramic X-ray image datasets. ResultsThe results of the three comparison experiments are shown that Accuracy, Precision, Recall, Dice, Volumetric Overlap Error and Relative Volume Difference for dental panoramic X-ray teeth segmentation are 98.53%, 95.62%, 94.51%, 94.28%, 88.92% and 95.97% by the proposed model respectively. ConclusionThe proposed modules complement each other in processing every detail of the dental panoramic X-ray images, which can effectively improve the efficiency of preoperative preparation and postoperative evaluation, and promote the application of dental panoramic X-ray in medical image segmentation. There are more accuracy about Teeth U-Net than others model in dental panoramic X-ray teeth segmentation. That is very important to clinical doctors to cure in orthodontics and restorative dentistry.

Automatic liver tumor segmentation on multiphase computed tomography volume using SegNet deep neural network and K‐means clustering

AbstractLiver and liver tumor segmentations are essential in computer‐aided systems for diagnosing liver tumors. These systems must operate on multiphase computed tomography (CT) images instead of a single phase for accurate diagnosis for clinical applications. We have proposed a framework that can perform segmentation from quadriphasic CT data. The liver was segmented using a fine‐tuned SegNet model and the liver tumor was segmented using the K‐means clustering method coupled with a power‐law transformation‐based image enhancement technique. The best values for liver segmentation achieved were: Dice Coefficient = 96.46 ± 0.48%, Jaccard Index = 93.16 ± 0.89%, volumetric overlap error = 6.84 ± 0.89% and average symmetric surface distance = 0.59 ± 0.3 mm and the results for liver tumor delineation were Dice Coefficient = 85.07 ± 4.5%, Jaccard Index = 74.29 ± 6.8%, volumetric overlap error = 25.71 ± 6.8% and average symmetric surface distance = 1.14 ± 1.3 mm. The proposed liver segmentation method based on deep learning is fully automatic, robust, and effective for all phases. The image enhancement technique has shown promising results and aided in better liver tumor segmentation. The liver tumors were segmented satisfactorily; however, improvements concerning false positive reduction can further increase the accuracy.

High-resolution computed tomography diagnosis of pneumoconiosis complicated with pulmonary tuberculosis based on cascading deep supervision U-Net

ObjectivePulmonary tuberculosis can promote pneumoconiosis deterioration, leading to higher mortality. This study aims to explore the diagnostic value of the cascading deep supervision U-Net (CSNet) model in pneumoconiosis complicated with pulmonary tuberculosis. MethodsA total of 162 patients with pneumoconiosis treated in our hospital were collected as the research objects. Patients were randomly divided into a training set (n = 113) and a test set (n = 49) in proportion (7:3). Based on the high-resolution computed tomography (HRCT), the traditional U-Net, supervision U-Net (SNet), and CSNet prediction models were constructed. Dice similarity coefficients, precision, recall, volumetric overlap error, and relative volume difference were used to evaluate the segmentation model. The area under the receiver operating characteristic curve (AUC) value represents the prediction efficiency of the model. ResultsThere were no statistically significant differences in gender, age, number of positive patients, and dust contact time between patients in the training set and test set (P > 0.05). The segmentation results of CSNet are better than the traditional U-Net model and the SNet model. The AUC value of the CSNet model was 0.947 (95% CI: 0.900∼0.994), which was higher than the traditional U-Net model. ConclusionThe CSNet based on chest HRCT proposed in this study is superior to the traditional U-Net segmentation method in segmenting pneumoconiosis complicated with pulmonary tuberculosis. It has good prediction efficiency and can provide more clinical diagnostic value.

Early severity prediction of BPD for premature infants from chest X-ray images using deep learning: A study at the 28th day of oxygen inhalation

Background and objectiveBronchopulmonary dysplasia is a common respiratory disease in premature infants. The severity is diagnosed at the 56th day after birth or discharge by analyzing the clinical indicators, which may cause the delay of the best treatment opportunity. Thus, we proposed a deep learning-based method using chest X-ray images of the 28th day of oxygen inhalation for the early severity prediction of bronchopulmonary dysplasia in clinic. MethodsWe first adopted a two-step lung field extraction method by combining digital image processing and human-computer interaction to form the one-to-one corresponding image and label. The designed XSEG-Net model was then trained for segmenting the chest X-ray images, with the results being used for the analysis of heart development and clinical severity. Therein, Six-Point cardiothoracic ratio measurement algorithm based on corner detection was designed for the analysis of heart development; and the transfer learning of deep convolutional neural network models were used for the early prediction of clinical severities. ResultsThe dice and cross-entropy loss value of the training of XSEG-Net network reached 0.9794 and 0.0146. The dice, volumetric overlap error, relative volume difference, precision, and recall were used to evaluate the trained model in testing set with the result being 98.43 ± 0.39%, 0.49 ± 0.35%, 0.49 ± 0.35%, 98.67 ± 0.40%, and 98.20 ± 0.47%, respectively. The errors between the Six-Point cardiothoracic ratio measurement method and the gold standard were 0.0122 ± 0.0084. The deep convolutional neural network model based on VGGNet had the promising prediction performance, with the accuracy, precision, sensitivity, specificity, and F1 score reaching 95.58 ± 0.48%, 95.61 ± 0.55%, 95.67 ± 0.44%, 96.98 ± 0.42%, and 95.61±0.48%, respectively. ConclusionsThese experimental results of the proposed methods in lung field segmentation, cardiothoracic ratio measurement and clinic severity prediction were better than previous methods, which proved that this method had great potential for clinical application.

Medical lesion segmentation by combining multimodal images with modality weighted UNet.

Automatic segmentation of medical lesions is a prerequisite for efficient clinic analysis. Segmentation algorithms for multimodal medical images have received much attention in recent years. Different strategies for multimodal combination (or fusion), such as probability theory, fuzzy models, belief functions, and deep neural networks, have also been developed. In this paper, we propose the modality weighted UNet (MW-UNet) and attention-based fusion method to combine multimodal images for medical lesion segmentation. MW-UNet is a multimodal fusion method which is based on UNet, but we use a shallower layer and fewer feature map channels to reduce the amount of network parameters, and our method uses the new multimodal fusion method called fusion attention. It uses weighted sum rule and fusion attention to combine feature maps in intermediate layers. During training, all the weight parameters are updated through backpropagation like other parameters in the network. We also incorporate residual blocks into MW-UNet to further improve segmentation performance. The comparison between the automatic multimodal lesion segmentations and the manual contours was quantified by (1) five metrics including Dice, 95% Hausdorff Distance (HD95), volumetric overlap error (VOE), relative volume difference (RVD), and mean-Intersection-over-Union (mIoU); (2) Number of parameters and flops to calculate the complexity of thenetwork. The proposed method is verified on ZJCHD, which is the data set of contrast-enhanced computed tomography (CECT) for Liver Lesion Segmentation taken from Cancer Hospital of the University of Chinese Academy of Sciences (Zhejiang Cancer Hospital), Hangzhou, China. For accuracy evaluation, we use 120 patients with liver lesions from ZJCHD, of which 100 are used for fourfold cross-validation (CV) and 20 are used for hold-out (HO) test. The mean Dice was and for HO and CV tests, respectively. The corresponding HD95, VOE, RVD, and mIoU of the two tests are 1.95 ± 1.83 and 2.67 ± 3.35 mm, 13.11 ± 15.83 and , 12.20 ± 18.20 and , and 83.79 ± 15.83 and . The parameters and flops of our method is 4.04 M and 18.36 G,respectively. The results show that our method performs well on multimodal liver lesion segmentation. It can be easily extended to other multimodal data sets and other networks for multimodal fusion. Our method is the potential to provide doctors with multimodal annotations and assist them with clinicaldiagnosis.

Using deep learning models to analyze the cerebral edema complication caused by radiotherapy in patients with intracranial tumor

Using deep learning models to analyze patients with intracranial tumors, to study the image segmentation and standard results by clinical depiction complications of cerebral edema after receiving radiotherapy. In this study, patients with intracranial tumors receiving computer knife (CyberKnife M6) stereotactic radiosurgery were followed using the treatment planning system (MultiPlan 5.1.3) to obtain before-treatment and four-month follow-up images of patients. The TensorFlow platform was used as the core architecture for training neural networks. Supervised learning was used to build labels for the cerebral edema dataset by using Mask region-based convolutional neural networks (R-CNN), and region growing algorithms. The three evaluation coefficients DICE, Jaccard (intersection over union, IoU), and volumetric overlap error (VOE) were used to analyze and calculate the algorithms in the image collection for cerebral edema image segmentation and the standard as described by the oncologists. When DICE and IoU indices were 1, and the VOE index was 0, the results were identical to those described by the clinician.The study found using the Mask R-CNN model in the segmentation of cerebral edema, the DICE index was 0.88, the IoU index was 0.79, and the VOE index was 2.0. The DICE, IoU, and VOE indices using region growing were 0.77, 0.64, and 3.2, respectively. Using the evaluated index, the Mask R-CNN model had the best segmentation effect. This method can be implemented in the clinical workflow in the future to achieve good complication segmentation and provide clinical evaluation and guidance suggestions.

Automatic Liver Tumor Segmentation in CT Modalities Using MAT-ACM

In the recent days, the segmentation of Liver Tumor (LT) has been demanding and challenging. The process of segmenting the liver and accurately spotting the tumor is demanding due to the diversity of shape, texture, and intensity of the liver image. The intensity similarities of the neighboring organs of the liver create difficulties during liver segmentation. The manual segmentation does not provide an accurate segmentation because the results provided by different medical experts can vary. Also, this manual technique requires a large number of image slices and time for segmentation. To solve these issues, the Fully Automatic Segmentation (FAS) technique is proposed. In this proposed Multi-Angle Texture Active Contour Model (MAT-ACM) method, the input Computed Tomography (CT) image is preprocessed by Contrast Enhancement (CE) with Non-Linear Mapping Technique (NLMT), in which the liver is differentiated from its neighbouring soft tissues with related strength. Then, the filtered images are given as the input to Adaptive Edge Modeling (AEM) with Canny Edge Detection (CED) technique, which segments the Liver Region (LR) from the given CT images. An AEM with a CED model is implemented, which increases the convergence speed of the iterative process for decreasing the Volumetric Overlap Error (VOE) is 6.92% rates when compared with the traditional Segmentation Techniques (ST). Finally, the Liver Tumor Segmentation (LTS) is developed by applying the MAT-ACM, which accurately segments the LR from the segmented LRs. The evaluation of the proposed method is compared with the existing LTS methods using various performance measures to prove the superiority of the proposed MAT-ACM method.

Liver segmentation from computed tomography images using cascade deep learning

BackgroundLiver segmentation is a fundamental step in the treatment planning and diagnosis of liver cancer. However, manual segmentation of liver is time-consuming because of the large slice quantity and subjectiveness associated with the specialist's experience, which can lead to segmentation errors. Thus, the segmentation process can be automated using computational methods for better time efficiency and accuracy. However, automatic liver segmentation is a challenging task, as the liver can vary in shape, ill-defined borders, and lesions, which affect its appearance. We aim to propose an automatic method for liver segmentation using computed tomography (CT) images. MethodsThe proposed method, based on deep convolutional neural network models and image processing techniques, comprise of four main steps: (1) image preprocessing, (2) initial segmentation, (3) reconstruction, and (4) final segmentation. ResultsWe evaluated the proposed method using 131 CT images from the LiTS image base. An average sensitivity of 95.45%, an average specificity of 99.86%, an average Dice coefficient of 95.64%, an average volumetric overlap error (VOE) of 8.28%, an average relative volume difference (RVD) of −0.41%, and an average Hausdorff distance (HD) of 26.60 mm were achieved. ConclusionsThis study demonstrates that liver segmentation, even when lesions are present in CT images, can be efficiently performed using a cascade approach and including a reconstruction step based on deep convolutional neural networks.

Adaptive Attention Convolutional Neural Network for Liver Tumor Segmentation.

PurposeAccurate segmentation of liver and liver tumors is critical for radiotherapy. Liver tumor segmentation, however, remains a difficult and relevant problem in the field of medical image processing because of the various factors like complex and variable location, size, and shape of liver tumors, low contrast between tumors and normal tissues, and blurred or difficult-to-define lesion boundaries. In this paper, we proposed a neural network (S-Net) that can incorporate attention mechanisms to end-to-end segmentation of liver tumors from CT images.MethodsFirst, this study adopted a classical coding-decoding structure to realize end-to-end segmentation. Next, we introduced an attention mechanism between the contraction path and the expansion path so that the network could encode a longer range of semantic information in the local features and find the corresponding relationship between different channels. Then, we introduced long-hop connections between the layers of the contraction path and the expansion path, so that the semantic information extracted in both paths could be fused. Finally, the application of closed operation was used to dissipate the narrow interruptions and long, thin divide. This eliminated small cavities and produced a noise reduction effect.ResultsIn this paper, we used the MICCAI 2017 liver tumor segmentation (LiTS) challenge dataset, 3DIRCADb dataset and doctors’ manual contours of Hubei Cancer Hospital dataset to test the network architecture. We calculated the Dice Global (DG) score, Dice per Case (DC) score, volumetric overlap error (VOE), average symmetric surface distance (ASSD), and root mean square error (RMSE) to evaluate the accuracy of the architecture for liver tumor segmentation. The segmentation DG for tumor was found to be 0.7555, DC was 0.613, VOE was 0.413, ASSD was 1.186 and RMSE was 1.804. For a small tumor, DG was 0.3246 and DC was 0.3082. For a large tumor, DG was 0.7819 and DC was 0.7632.ConclusionS-Net obtained more semantic information with the introduction of an attention mechanism and long jump connection. Experimental results showed that this method effectively improved the effect of tumor recognition in CT images and could be applied to assist doctors in clinical treatment.

PA-ResSeg: A phase attention residual network for liver tumor segmentation from multiphase CT images.

Liver tumor segmentation is a crucial prerequisite for computer-aided diagnosis of liver tumors. In the clinical diagnosis of liver tumors, radiologists usually examine multiphase CT images as these images provide abundant and complementary information of tumors. However, most known automatic segmentation methods extract tumor features from CT images merely of a single phase, in which valuable multiphase information is ignored. Therefore, it is highly demanded to develop a method effectively incorporating multiphase information for automatic and accurate liver tumor segmentation. In this paper, we propose a phase attention residual network (PA-ResSeg) to model multiphase features for accurate liver tumor segmentation. A phase attention (PA) is newly proposed to additionally exploit the images of arterial (ART) phase to facilitate the segmentation of portal venous (PV) phase. The PA block consists of an intraphase attention (intra-PA) module and an interphase attention (inter-PA) module to capture channel-wise self-dependencies and cross-phase interdependencies, respectively. Thus, it enables the network to learn more representative multiphase features by refining the PV features according to the channel dependencies and recalibrating the ART features based on the learned interdependencies between phases. We propose a PA-based multiscale fusion (MSF) architecture to embed the PA blocks in the network at multiple levels along the encoding path to fuse multiscale features from multiphase images. Moreover, a 3D boundary-enhanced loss (BE-loss) is proposed for training to make the network more sensitive to boundaries. To evaluate the performance of our proposed PA-ResSeg, we conducted experiments on a multiphase CT dataset of focal liver lesions (MPCT-FLLs). Experimental results show the effectiveness of the proposed method by achieving a dice per case (DPC) of 0.7787, a dice global (DG) of 0.8682, a volumetric overlap error (VOE) of 0.3328, and a relative volume difference (RVD) of 0.0443 on the MPCT-FLLs. Furthermore, to validate the effectiveness and robustness of PA-ResSeg, we conducted extra experiments on another multiphase liver tumor dataset and obtained a DPC of 0.8290, a DG of 0.9132, a VOE of 0.2637, and a RVD of 0.0163. The proposed method shows its robustness and generalization capability in different datasets and different backbones. The study demonstrates that our method can effectively model information from multiphase CT images to segment liver tumors and outperforms other state-of-the-art methods. The PA-based MSF method can learn more representative multiphase features at multiple scales and thereby improve the segmentation performance. Besides, the proposed 3D BE-loss is conducive to tumor boundary segmentation by enforcing the network focus on boundary regions and marginal slices. Experimental results evaluated by quantitative metrics demonstrate the superiority of our PA-ResSeg over the best-known methods.

An automatic method for segmentation of liver lesions in computed tomography images using deep neural networks

Liver cancer is one of the major causes of death by cancer. The early detection of lesions in the liver provides a better chance of treatment and cure of the disease. Computed tomography (CT) is one of the most used imaging techniques for the detection and diagnosis of liver lesions. However, the manual segmentation of liver and tumors, aside from being time-consuming, can still cause errors and may vary among specialists. Because of this hard work, computer-aided detection (CAD) and computer-aided diagnosis (CADx) systems have been developed to assist specialists in the detection and characterization of lesions in the liver and reduce the required time for diagnosis. The automatic segmentation of these lesions is a complex task since they present variability in contrast, shape, size, and location. In this work, a method to automatically segment liver lesions in CT images is proposed. The proposed method, which presents two deep convolutional neural networks (CNN) models, consists of five main steps: (1) image acquisition, (2) image pre-processing, (3) initial segmentation using RetinaNet, (4) lesion segmentation using U-Net, and (5) segmentation refinement. The proposed method was evaluated using a set of 131 CT images from the LiTS dataset, and the best result obtained a matthews correlation coefficient (MCC) of 83.62%, a sensitivity of 83.86%, a specificity of 99.96%, a Dice coefficient of 82.99%, a volumetric overlap error (VOE) of 27.89%, and a relative volume difference (RVD) of 1.69%. We show in our method that the problem of segmentation of liver lesions in CT images can be efficiently solved through the use of deep CNNs to define the scope of the problem and to precisely segment lesions.

Automated cartilage segmentation and quantification using 3D ultrashort echo time (UTE) cones MR imaging with deep convolutional neural networks.

To develop a fully automated full-thickness cartilage segmentation and mapping of T1, T1ρ, and T2*, as well as macromolecular fraction (MMF) by combining a series of quantitative 3D ultrashort echo time (UTE) cones MR imaging with a transfer learning-based U-Net convolutional neural networks (CNN) model. Sixty-five participants (20 normal, 29 doubtful-minimal osteoarthritis (OA), and 16 moderate-severe OA) were scanned using 3D UTE cones T1 (Cones-T1), adiabatic T1ρ (Cones-AdiabT1ρ), T2* (Cones-T2*), and magnetization transfer (Cones-MT) sequences at 3T. Manual segmentation was performed by two experienced radiologists, and automatic segmentation was completed using the proposed U-Net CNN model. The accuracy of cartilage segmentation was evaluated using the Dice score and volumetric overlap error (VOE). Pearson correlation coefficient and intraclass correlation coefficient (ICC) were calculated to evaluate the consistency of quantitative MR parameters extracted from automatic and manual segmentations. UTE biomarkers were compared among different subject groups using one-way ANOVA. The U-Net CNN model provided reliable cartilage segmentation with a mean Dice score of 0.82 and a mean VOE of 29.86%. The consistency of Cones-T1, Cones-AdiabT1ρ, Cones-T2*, and MMF calculated using automatic and manual segmentations ranged from 0.91 to 0.99 for Pearson correlation coefficients, and from 0.91 to 0.96 for ICCs, respectively. Significant increases in Cones-T1, Cones-AdiabT1ρ, and Cones-T2* (p<0.05) and a decrease in MMF (p<0.001) were observed in doubtful-minimal OA and/or moderate-severe OA over normal controls. Quantitative 3D UTE cones MR imaging combined with the proposed U-Net CNN model allows a fully automated comprehensive assessment of articular cartilage. • 3D UTE cones imaging combined with U-Net CNN model was able to provide fully automated cartilage segmentation. • UTE parameters obtained from automatic segmentation were able to reliably provide a quantitative assessment of cartilage.