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.

No abstract

This paper investigates the optimal training duration and the optimal power allocation for the training and the data transmission that maximize the ergodic sum rate in single-cell uplink massive MIMO with MMSE receivers. Our channel model assumes that each user experiences the same spatial channel correlation. The expression for the ergodic sum rate is obtained in the large system regime where the number of antennas (N) at the base station and the number of users (K) tend to infinity with a fixed ratio. Interestingly, we show that the optimal training duration is equal to K and independent of the spatial correlation. We also derive the optimal power allocation that in facts depends on the spatial correlation, the channel coherence interval, and the uplink SNR. We show that more energy should be allocated for the training if the data transmission duration (t d) is less than K , and vice versa. Moreover, equal power allocation is optimal when t d = K. We also obtain an approximation for the optimal power allocation that depends on the mean of the correlation matrix eigenvalues. Numerical simulations show that our results based on the large system approximation are accurate and applicable for finite-size systems. The simulations also show that (1) the resulting ergodic sum rates obtained by employing the optimal power allocation and its approximation are indistinguishable, and (2) the optimal power allocation obtained from the uncorrelated channel model can be applied to the cases involving the correlated channels with indiscernible penalties on the ergodic sum rates.

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.

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.