International Journal of Imaging Systems and Technology最新文献

Glioma is a type of brain tumor with a high mortality rate. Magnetic resonance imaging (MRI) is commonly used for examination, and the accurate segmentation of tumor regions from MR images is essential to computer-aided diagnosis. However, due to the intrinsic heterogeneity of brain glioma, precise segmentation is very challenging, especially for tumor subregions. This article proposed a two-stage cascaded method for brain tumor segmentation that considers the hierarchical structure of the target tumor subregions. The first stage aims to identify the whole tumor (WT) from the background area; and the second stage aims to achieve fine-grained segmentation of the subregions, including enhanced tumor (ET) region and tumor core (TC) region. Both stages apply a deep neural network structure combining modified 3D U-Net with a residual connection scheme to tumor region and subregion segmentation. Moreover, in the training phase, the 3D masks generation of subregions with potential incomplete connectivity are guided by the completely connected regions. Experiments were performed to evaluate the performance of the methods on both area and boundary accuracy. The average dice score of the WT, TC, and ET regions on BraTS 2020 dataset is 0.9168, 0.0.8992, 0.8489, and the Hausdorff distance is 6.021, 9.203, 12.171, respectively. The proposed method outperforms current works, especially in segmenting fine-grained tumor subregions.

Diabetic retinopathy is a complication of diabetes and one of the leading causes of vision loss. Early detection and treatment are essential to prevent vision loss. Deep learning has been making great strides in the field of medical image processing and can be used as an aid for medical practitioners. However, unbalanced datasets, sparse focal areas, small differences between adjacent disease grades, and varied manifestations of the same grade disease challenge deep learning model training. Generalization performance and robustness are inadequate. To address the problem of unbalanced sample numbers between classes in the dataset, this work proposes using VQ-VAE for reconstructing affine transformed images to enrich and balance the dataset. Test results show the model's average reconstruction error is 0.0001, and the mean structural similarity between reconstructed and original images is 0.967. This proves reconstructed images differ from originals yet belong to the same category, expanding and diversifying the dataset. Addressing the issues of focal area sparsity and disease grade disparity, this work utilizes ResNeXt50 as the backbone network and constructs diverse attention networks by modifying the network structure and embedding different attention modules. Experiments demonstrate that the convolutional attention network outperforms ResNeXt50 in terms of Precision, Sensitivity, Specificity, F1 Score, Quadratic Weighted Kappa Coefficient, Accuracy, and robustness against Salt and Pepper noise, Gaussian noise, and gradient perturbation. Finally, the heat maps of each model recognizing the fundus image were plotted using the Grad-CAM method. The heat maps show that the attentional network is more effective than the non-attentional network ResNeXt50 at attending to the fundus image.

Accurate segmentation of the optic cup and disc in fundus images is crucial for the prevention and diagnosis of glaucoma. However, challenges arise due to factors such as blood vessels, and mainstream networks often demonstrate limited capacity in extracting contour information. In this paper, we propose a segmentation framework named FDT-Net, which is based on a frequency-aware dual-branch Transformer (FDBT) architecture with parallel contour information mining and uncertainty-guided refinement. Specifically, we design a FDBT that operates in the frequency domain. This module leverages the inherent contextual awareness of Transformers and utilizes Discrete Cosine Transform (DCT) transformation to mitigate the impact of certain interference factors on segmentation. The FDBT comprises global and local branches that independently extract global and local information, thereby enhancing segmentation results. Moreover, to further mine additional contour information, this study develops the parallel contour information mining (PCIM) module to operate in parallel. These modules effectively capture more details from the edges of the optic cup and disc while avoiding mutual interference, thus optimizing segmentation performance in contour regions. Furthermore, we propose an uncertainty-guided refinement (UGR) module, which generates and quantifies uncertainty mass and leverages it to enhance model performance based on subjective logic theory. The experimental results on two publicly available datasets demonstrate the superior performance and competitive advantages of our proposed FDT-Net. The code for this project is available at https://github.com/Rookie49144/FDT-Net.

Skin cancer is a common malignant tumor causing tens of thousands of deaths each year, making early detection essential for better treatment outcomes. However, the similar visual characteristics of skin lesions make it challenging to accurately differentiate between lesion types. With advancements in deep learning, researchers have increasingly turned to convolutional neural networks for skin cancer detection and classification. In this article, an improved skin cancer classification model M-Net is proposed, and the enhanced gray wolf optimization algorithm is combined to improve the classification performance. The gray wolf optimization algorithm guides the wolf pack to prey through a multileader structure and gradually converges through the encirclement and pursuit mechanism, so as to perform a more detailed search in the later stage. To further improve the performance of the gray wolf optimization, this study introduces the simulated annealing algorithm to avoid falling into the local optimal state and expands the search range by improving the search mechanism, thus enhancing the global optimization ability of the algorithm. The M-Net model significantly improves the accuracy of classification by extracting features of skin lesions and optimizing parameters with the enhanced gray wolf optimization algorithm. The experimental results based on the ISIC 2018 dataset show that compared with the baseline model, the feature extraction network of the model has achieved a significant improvement in accuracy. The classification performance of M-Net is excellent in multiple indicators, with accuracy, precision, recall, and F1 score reaching 0.891, 0.857, 0.895, and 0.872, respectively. In addition, the modular design of M-Net enables it to flexibly adjust feature extraction and classification modules to adapt to different classification tasks, showing great scalability and applicability. In general, the model proposed in this article performs well in the classification of skin lesions, has broad clinical application prospects, and provides strong support for promoting the diagnosis of skin diseases.

Throughout the past 20 years, medical imaging has found extensive application in clinical diagnosis. Doctors may find it difficult to diagnose diseases using only one imaging modality. The main objective of multimodal medical image fusion (MMIF) is to improve both the accuracy and quality of clinical assessments by extracting structural and spectral information from source images. This study proposes a novel MMIF method to assist doctors and postoperations such as image segmentation, classification, and further surgical procedures. Initially, the intensity-hue-saturation (IHS) model is utilized to decompose the positron emission tomography (PET)/single photon emission computed tomography (SPECT) image, followed by a hue-angle mapping method to discriminate high- and low-activity regions in the PET images. Then, a proposed structure feature adjustment (SFA) mechanism is used as a fusion strategy for high- and low-activity regions to obtain structural and anatomical details with minimum color distortion. In the second step, a new multi-discriminator generative adversarial network (MDcGAN) approach is proposed for obtaining the final fused image. The qualitative and quantitative results demonstrate that the proposed method is superior to existing MMIF methods in preserving the structural, anatomical, and functional details of the PET/SPECT images. Through our assessment, involving visual analysis and subsequent verification using statistical metrics, it becomes evident that color changes contribute substantial visual information to the fusion of PET and MR images. The quantitative outcomes demonstrate that, in the majority of cases, the proposed algorithm consistently outperformed other methods. Yet, in a few instances, it achieved the second-highest results. The validity of the proposed method was confirmed using diverse modalities, encompassing a total of 1012 image pairs.

As an emerging technology that uses a pill-sized camera to visualize images of the digestive tract. Wireless capsule endoscopy (WCE) presents several advantages, since it is far less invasive, does not need sedation and has less possible complications compared to standard endoscopy. Hence, it might be exploited as alternative to the standard procedure. WCE is used to diagnosis a variety of gastro-intestinal diseases such as polyps, ulcers, crohns disease, and hemorrhages. Nevertheless, WCE videos produced after a test may consist of thousands of frames per patient that must be viewed by medical specialists, besides, the capsule free mobility and technological limits cause production of a low quality images. Hence, development of an automatic tool based on artificial intelligence might be very helpful. Moreover, most state-of-the-art works aim at images classification (normal/abnormal) while ignoring diseases segmentation. Therefore, in this work a novel method based on Feature Pyramid Network model is presented. This approach aims at diseases segmentation from WCE images. In this model, modules to optimize and combine features were employed. Specifically, semantic and spatial features were mutually compensated by spatial attention and cross-level global feature fusion modules. The proposed method testing F1-score and mean intersection over union are 94.149% and 89.414%, respectively, in the MICCAI 2017 dataset. In the KID Atlas dataset, the method achieved a testing F1-score and mean intersection over union of 94.557% and 90.416%, respectively. Through the performance analysis, the mean intersection over union in the MICCAI 2017 dataset is 20.414%, 18.484%, 11.444%, 8.794% more than existing approaches. Moreover, the proposed scheme surpassed the methods used for comparison by 29.986% and 9.416% in terms of mean intersection over union in KID Atlas dataset. These results indicate that the proposed approach is promising in diseases segmentation from WCE images.

The healthcare industry has been significantly impacted by the widespread adoption of advanced technologies such as deep learning (DL) and artificial intelligence (AI). Among various applications, computer-aided diagnosis has become a critical tool to enhance medical practice. In this research, we introduce a hybrid approach that combines a deep neural model, data collection, and classification methods for CT scans. This approach aims to detect and classify the severity of pulmonary disease and the stages of lung cancer. Our proposed lung cancer detector and stage classifier (LCDSC) demonstrate greater performance, achieving higher accuracy, sensitivity, specificity, recall, and precision. We employ an active contour model for lung cancer segmentation and high-resolution net (HRNet) for stage classification. This methodology is validated using the industry-standard benchmark image dataset lung image database consortium and image database resource initiative (LIDC-IDRI). The results show a remarkable accuracy of 98.4% in classifying lung cancer stages. Our approach presents a promising solution for early lung cancer diagnosis, potentially leading to improved patient outcomes.

In recent years, U-Net and its variants have gained widespread use in medical image segmentation. One key aspect of U-Net's design is the skip connection, facilitating the retention of detailed information and leading to finer segmentation results. However, existing research often concentrates on enhancing either the encoder or decoder, neglecting the semantic gap between them, and resulting in suboptimal model performance. In response, we introduce Multi-Scale Fusion module aimed at enhancing the original skip connections and addressing the semantic gap. Our approach fully incorporates the correlation between outputs from adjacent encoder layers and facilitates bidirectional information exchange across multiple layers. Additionally, we introduce Channel Relation Perception module to guide the fused multi-scale information for efficient connection with decoder features. These two modules collectively bridge the semantic gap by capturing spatial and channel dependencies in the features, contributing to accurate medical image segmentation. Building upon these innovations, we propose a novel network called MFH-Net. On three publicly available datasets, ISIC2016, ISIC2017, and Kvasir-SEG, we perform a comprehensive evaluation of the network. The experimental results show that MFH-Net exhibits higher segmentation accuracy in comparison with other competing methods. Importantly, the modules we have devised can be seamlessly incorporated into various networks, such as U-Net and its variants, offering a potential avenue for further improving model performance.