No abstract

Developing a breast cancer screening method is very important to facilitate early breast cancer detection and treatment. Building a screening method using medical imaging modality that does not cause body tissue damage (non-invasive) and does not involve physical touch is challenging. Thermography, a non-invasive and non-contact cancer screening method, can detect tumors at an early stage even under precancerous conditions by observing temperature distribution in both breasts. The thermograms obtained on thermography can be interpreted using deep learning models such as convolutional neural networks (CNNs). CNNs can automatically classify breast thermograms into categories such as normal and abnormal. Despite their demostrated utility, CNNs have not been widely used in breast thermogram classification. In this study, we aimed to summarize the current work and progress in breast cancer detection based on thermography and CNNs. We first discuss of breast thermography potential in early breast cancer detection, providing an overview of the availability of breast thermal datasets together with publicly accessible. We also discuss characteristics of breast thermograms and the differences between healthy and cancerous thermographic patterns. Breast thermogram classification using a CNN model is described step by step including a simulation example illustrating feature learning. We cover most research related to the implementation of deep neural networks for breast thermogram classification and propose future research directions for developing representative datasets, feeding the segmented image, assigning a good kernel, and building a lightweight CNN model to improve CNN performance. INDEX TERMS breast cancer; convolutional neural network; deep learning; early detection; thermogram

We consider a multiuser system where a single transmitter equipped with multiple antennas (the base station) communicates with multiple users each with a single antenna. Regularized channel inversion is employed as the precoding strategy at the base station. Within this scenario we are interested in the problems of power allocation and user admission control so as to maximize the system throughput, i.e., which users should we communicate with and what power should we use for each of the admitted users so as to get the highest sum rate. This is in general a very difficult problem but we do two things to allow some progress to be made. Firstly we consider the large system regime where the number of antennas at the base station is large along with the number of users. Secondly we cluster the downlink path gains of users into a finite number of groups. By doing this we are able to show that the optimal power allocation under an average transmit power constraint follows the well-known water filling scheme. We also investigate the user admission problem which reduces in the large system regime to optimization of the user loading in the system. Index Terms-Multiuser precoding, regularized channel inversion, power allocation, large system analysis.

In this paper, we study feedback optimization problems that maximize the users' signal to interference plus noise ratio (SINR) in a two-cell MIMO broadcast channel. Assuming the users learn their direct and interfering channels perfectly, they can feed back this information to the base stations (BSs) over the uplink channels. TheBSs then use the channel information to design their transmission scheme. Two types of feedback are considered: analog and digital. In the analog feedback case, the users send their unquantized and uncoded CSI over the uplink channels. In this context, given a user's fixed transmit power, we investigate how he/she should optimally allocate it to feed back the direct and interfering (or cross) CSI for two types of base station cooperation schemes, namely, Multi-Cell Processing (MCP) and Coordinated Beamforming (CBf). In the digital feedback case, the direct and cross link channel vectors of each user are quantized separately, each using RVQ, with different size codebooks.The users then send the index of the quantization vector in the corresponding codebook to the BSs. Similar to the feedback optimization problem in the analog feedback, we investigate the optimal bit partitioning for the direct and interfering link for both types of cooperation.We focus on regularized channel inversion precoding structures and perform our analysis in the large system limit in which the number of users per cell (K) and the number of antennas per BS (N ) tend to infinity with their ratio β = K N held fixed. We show that for both types of cooperation, for some values of interfering channel gain, usually at low values, no cooperation between the base stations is preferred: This is because, for these values of cross channel gain, the channel estimates for the cross link are not accurate enough for their knowledge to contribute to improving the SINR and there is no benefit in doing base station cooperation under that condition. We also show that for the MCP scheme, unlike in the perfect CSI case, the SINR improves only when the interfering channel gain is above a certain threshold.

Early detection of plant diseases is one of the main keys to handling diseases quickly and successfully. The purpose of this study is to find out a simpler CNN architecture and meet an acceptable compromise between accuracy and simplification to detect diseases in tomato plants based on leaf images. This simpler architecture will allow the development of standalone and independent system model in the field to classify and identify the tomato plants diseases in low price and limited resources. This proposed architecture was developed from the CNN architecture baseline and is intended to classify 10 classes of tomato leaves consist of one healthy class and 9 classes of leaves diseases taken from the Plant Village dataset. In this study, the performance of the proposed architecture and comparative architectures are examined in the same dataset. Comparative architectures used are some existing CNN architectures that are commonly used namely VGG Net, Shuffle Net and Squeeze Net. The results indicated that the proposed architecture can achieve competitive accuracy compared with the existing architecture while the proposed architecture is much shorter than the existing architecture and better in terms of performance time.

In general, most of the existing convolutional neural network (CNN)-based deep-learning models suffer from spatial-information loss and inadequate feature-representation issues. This is due to their inability to capture multiscale-context information and the exclusion of semantic information throughout the pooling operations. In the early layers of a CNN, the network encodes simple semantic representations, such as edges and corners, while, in the latter part of the CNN, the network encodes more complex semantic features, such as complex geometric shapes. Theoretically, it is better for a CNN to extract features from different levels of semantic representation because tasks such as classification and segmentation work better when both simple and complex feature maps are utilized. Hence, it is also crucial to embed multiscale capability throughout the network so that the various scales of the features can be optimally captured to represent the intended task. Multiscale representation enables the network to fuse low-level and high-level features from a restricted receptive field to enhance the deep-model performance. The main novelty of this review is the comprehensive novel taxonomy of multiscale-deep-learning methods, which includes details of several architectures and their strengths that have been implemented in the existing works. Predominantly, multiscale approaches in deep-learning networks can be classed into two categories: multiscale feature learning and multiscale feature fusion. Multiscale feature learning refers to the method of deriving feature maps by examining kernels over several sizes to collect a larger range of relevant features and predict the input images’ spatial mapping. Multiscale feature fusion uses features with different resolutions to find patterns over short and long distances, without a deep network. Additionally, several examples of the techniques are also discussed according to their applications in satellite imagery, medical imaging, agriculture, and industrial and manufacturing systems.

No abstract

The designs of Islamic women apparels is dynamically changing, which can be shown by emerging of online shops selling clothing with fast updates of newest models. Traditionally, buying the clothes online can be done by querying the keywords to the retrieval system. The approach has a drawback that the keywords cannot describe the clothes designs precisely. Therefore, a searching based on content–known as content-based image retrieval (CBIR)–is required. One of the features used in CBIR is the shape. This article presents a new normalization approach to the Pyramid Histogram of Oriented Gradients (PHOG) as a mean for shape feature extraction of women Islamic clothing in a retrieval system. We refer to the proposed approach as normalized PHOG (NPHOG). The Euclidean distance measured the similarity of the clothing. The performance of the system was evaluated by using 340 clothing images, comprised of four clothing categories, 85 images for each category: blouse-pants, long dress, outerwear, and tunic. The recall and precision parameters measured the retrieval performance; the Histogram of Oriented Gradients (HOG) and PHOG were the methods for comparison. The experiments showed that NPHOG improved the HOG and PHOG performance in three clothing categories.