Search

Mammography technique is commonly used for diagnosing breast cancer, but mammography images usually show low contrast which cause difficulties to clinical diagnosis. Therefore, improving the visual quality of mammography images is an important issue. This is a challenging problem, because every mammography image consists of rich textures, including bright areas, dark areas, and textural details. Inspired by bio-inspired neural network, this paper proposes a Bidirectional Spiking Cortical Model (Bi-SCM) from the perspective of neural information fusion to enhance the contrast of bright areas and dark areas adequately, as well as textural details. This goal is achieved by utilizing the Bi-SCM to first enhance a mammography image and its inverse separately. The enhanced results are fused by a new fusion algorithm based on non-subsampled contourlet transform (NSCT) to ensure that both of the contrast of bright areas and dark areas are adequately improved. The textual details are then enhanced by an unsharp masking method which consists of cubic filter and log-ratio operation. Sufficient experiments on mammography images are conducted to evaluate the proposed approach. Experimental results show that the proposed method outperforms state-of-the-art methods on both enhancing contrast and details. Besides, over-enhancement and noise sensitivity are also significantly suppressed. [ABSTRACT FROM AUTHOR]

Pulse-coupled Neural Network (PCNN) is a third-generation artificial neural network that requires no training. Neurons in PCNN have two pulse burst modes: firing mode and fire-extinguishing mode. A lot of research has been conducted on achieving image segmentation using the fire-extinguishing mode, yet remains deficient on the characteristics and applications of the firing mode. Through analysis of the firing process of PCNN, we find that the network that works only on the firing mode has the characteristic of image edge detection. Then, we give a mathematical expression for the neuron firing time, and, using the expression, propose an image edge detection algorithm based on the characteristic of PCNN double step firing. Then, we analyze the parameter constraints for achieving PCNN double step firing and provide a method for the self-adaptive setting of the network parameters. To achieve the best edge detection, we conduct mathematical analysis on the relationship between edge detection performance and neighbor coupling, and provide a neighbor template structure and the setting of its values. Our results show that the proposed algorithm can obtain smooth and unbroken single pixels edges, and has nice robustness and efficiency for various kinds of images. [ABSTRACT FROM AUTHOR]

Intra-class inconsistency and inter-class indistinction are intractable problems that commonly exist in breast mass segmentation from mammograms. In this work, a novel deep learning segmentation model is presented to address these problems. Firstly, we propose a simple yet effective aggregated pyramid attention module (APAM) for capturing intra-class dependencies, aiming at effectively aggregating contextual dependencies from different receptive fields to reinforce feature representations. Then, a novel aggregated pyramid attention network (APANet) is developed for further releasing the limitation of both intra-class inconsistency and inter-class indistinction. The APANet can combine low-level spatial details and high-level contextual information via encoder-decoder structure for further refining semantic representations. Finally, our proposed APANet is greatly demonstrated on two public mammographic databases including the DDSM-BCRP and INbreast, separately achieving the Dice Similarity Coefficient (DSC) of 91.04% and 94.02%. [ABSTRACT FROM AUTHOR]

Automated benign and malignant breast masses classification is a crucial yet challenging topic. Recently, many studies based on convolutional neural network (CNN) are presented to address this task, but most of these CNN-based methods neglect the effective global contextual information. Moreover, their methods do not further analyze the reliability and interpretability of CNN models, which does not correspond to the clinical diagnosis. In this work, we firstly propose a novel multi-level global-guided branch-attention network (MGBN) for mass classification, which aims to fully leverage the multi-level global contextual information to refine the feature representation. Specifically, the MGBN includes a stem module and a branch module. The former extracts the local information through standard local convolutional operations of ResNet-50. The latter embeds the global contextual information and establishes the relationships of different feature levels via global pooling and Multi-layer Perceptron (MLP). The final prediction is computed by local information and global information together. Then, we discuss the reliability and interpretability of our mass classification network by visualizing the coarse localization map through Gradient-weighted Class Activation Mapping (Grad-CAM), which is important in clinical diagnosis. Finally, our proposed MGBN is greatly demonstrated on two public mammographic mass classification databases including the DDSM and INbreast databases, resulting in AUC of 0.8375 and 0.9311, respectively. [ABSTRACT FROM AUTHOR]

Breast density is known as a significant indicator of breast cancer risk prediction and greatly reduces the digital mammograms sensitivity. In this work, based on the simple pulse coupled neural network (SPCNN), a novel Morph_SPCNN model is proposed for dealing with the limitations of over-segmentation that commonly existed in density segmentation of mammograms. To evaluate the proposed model, the segmentation result is employed as a feature map of the breast density classification system. In addtion, the texture features of mammogram calculated based on the gray level co-occurrence matrix (GLCM) and the statistical features (mean, skewness, kurtosis) are extracted and input to the support vector machine (SVM) for breast density classification. Finally, the performance of SVM classifier is evaluated based on the ten-fold cross-validation. Our method is verified both on the MIAS dataset, DDSM database and hybrid dataset (MIAS database and Gansu Provincial Academy of Medical Sciences (GPAMS) database), respectively achieving 87.80%, 94.89% and 95.37% accuracy for breast density classification. The experimental results indicate that our proposed method has greatly improved the performance of breast density segmentation and classification. [ABSTRACT FROM AUTHOR]

Breast cancer poses a threat to the lives of many women. Breast density is a closely related indicator of breast cancer risk. The aim of this paper is to propose a classification system for breast density, which can appropriately segment the glandular tissue from the whole breast and to achieve a better classification result. A new threshold method is applied to segment the breast glandular tissue. The gray level co-occurrence matrix (GLCM) is implemented to extract the texture features of the glandular tissue. Meanwhile, we obtain three statistical features (mean, skewness, kurtosis). In addition, the calculated breast density that is served as a new feature is added to the feature vectors. The mixed feature vectors are classified by Support Vector Machine (SVM) and Ultimate Learning Machine (ELM). Ten-fold cross-validation is used to verify the classifier performance. The system using the SVM achieves 96.19% accuracy for three density types in the MIAS database and achieves 96.35% accuracy of four density types in the DDSM database. The accuracy in the database mixed with the local database was 95.01% and there are three density types in the mixed database. The experimental results indicate that the system proposed has a better performance in breast density classification. The system proposed in this paper can be considered to help the physician to classify breast density. [ABSTRACT FROM AUTHOR]

Nowadays, embedded systems can be found everywhere in daily life. In the development of embedded systems, data security is one of the critical factors. Encryption is an effective way to protect data from threats. Among encryption algorithms, chaos-based methods have strong cryptographic properties since chaotic systems are sensitive to initial conditions and parameters. However, most of these algorithms cannot be applied in practice because their encryption speed is not fast enough. In this paper, a fast image encryption algorithm is proposed. Compared with traditional chaos-based image encryption algorithms, the proposed method utilizes mixed-sequence and decorrelation operation to enhance the randomness of chaotic sequence. Moreover, it used minimum length of the sequence which is determined by experiments. Therefore, the proposed scheme spends much less computation time, which is an important advantage for being applied in practice. Testing results have shown that this algorithm has good performance in resisting known attacks, such as known-plaintext attacks, chosen ciphertext attacks, statistical attacks, differential attacks, and various brute-force attacks. [ABSTRACT FROM AUTHOR]

A novel approach for mass segmentation from coarse-to-fine in digital mammography, termed as SCM-motivated enhanced CV algorithm, is presented in this paper. As well known, it is difficult to robustly achieve mammogram mass segmentation due to low contrast between normal and lesion tissues, as well as high density tissue interference in mammograms. Therefore, Spiking Cortical Model with biology background is introduced to achieve mammary-specific and mass edge detection, and this mass candidate is regarded as the initial contour of improved CV model followed by, effectively overcoming the drawback that CV method is sensitive to the initial contour; especially, the enhanced CV model innovatively combines the techniques of physical imaging principle, and local region-scalable force, harvesting the coarse-to-fine mass boundary accurately. The proposed method is tested totally on 400 mammograms from two well-known digitized datasets (digital database for screening mammography and mammography image analysis society database), achieving the average detection rate of 93.25%. By comparing with the region-based model with bias field (Method 1) and typical CV model (Method 2), we can reach the conclusion that proposed method is outperform other methods, yielding the average sensitivity of 95.83%, specificity of 99.13%, dice similarity co-efficient of 92.21% and AUC of 98.02%. In addition, this method is verified on the mammograms from Gansu Provincial Cancer Hospital, the detection results reveal that our method can accurately detect the abnormal in clinical application. [ABSTRACT FROM AUTHOR]

Today, medical imaging suffers from serious issues such as malicious tampering and privacy leakage. Encryption is an effective way to protect these images from security threats. Among the available encryption algorithms, chaos-based methods have strong cryptographic properties, because chaotic systems are sensitive to initial conditions and parameters. However, traditional chaotic systems are easy to build, analyze, predict and can be re-scaled to any desired frequency. Thus, encryption schemes using traditional chaotic systems have low security levels. In this work, we propose a new simple chaotic system that utilizes a hyperbolic sine as its nonlinearity; this nonlinearity has rarely appeared in previous studies. Furthermore, the new chaotic system uses a decorrelation operation to enhance its performance. Statistical testing verifies that the chaotic sequence has good pseudorandom characteristics. In this study, we propose a scheme for medical image encryption based on this new chaotic system. The results of tests show that this encryption method can encrypt images effectively in a single round and that the proposed scheme provides sufficient security against known attacks. [ABSTRACT FROM AUTHOR]

Local structural information is crucially important for human visual system to perceive natural scenes. Recent years, a variety of local image structure description methods have been proposed for the artificial modeling of visual perception. Although existing local image structure descriptors have shown successful performances, one general limitation is their numerical instability caused by ignoring the information of the spatial correlation of local orientation. In this paper, we propose a local image structure descriptor by modeling the anisotropic mechanism in the primary visual cortex. In particular, the pixel-wise anisotropy values of a given image are calculated by pseudo-Wigner-Ville distribution (PWVD) and Rényi entropy. Then the excitatory/inhibitory interactions among visual neurons in the local receptive field are modeled by measuring the similarities between their anisotropies. By mapping visual neurons to image pixels, the correlation between a central pixel and its local neighbors can be represented by a binary pattern which is named as local anisotropic pattern (LAP). Experimental results on texture classification verified that the proposed LAP has satisfactory texture classification accuracy, rotation invariance, and noise robustness; experimental results on no-reference image quality assessment demonstrated that the proposed LAP achieves state-of-the-art performance in objective evaluation of the perceptual quality of natural image, and this reflects that LAP can accurately represent the degradation of local image structure. [ABSTRACT FROM AUTHOR]

Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. The online version of the original article can be found at https://doi.org/10.1007/s11042-022-12725-2 B Correction to: Multimedia Tools and Applications b https://doi.org/10.1007/s11042-022-12725-2 The citations for reference [46] were missing in the original publication of this article. [Extracted from the article]