Search

Researchers' choice of metrics and criteria in evaluating recommender systems depends on what the researcher feels is popular among other researchers, or sometimes based on the objective of the research. There is no harmonized set of criteria and metrics that can be referenced when evaluating recommender systems in healthcare. In this article, a set of metrics and criteria are harmonized and categorized as a guide for evaluating recommender systems. By means of an online survey, the opinions of forty-four experienced researchers and other stakeholders from eight countries and four continents were sought on the relevance of identified metrics and criteria. Analysis of the results show speed and timeliness are at the top. Topping the list of criteria is the provision of information that will guide users to useful decisions. The result is presented from two logical perspectives. Four categories are then identified as a useful guide for evaluating recommender systems.

PurposeThe roles recommendation systems play in health care have become crucial in achieving effective care and in meeting the needs of modern care giving. As a result, efforts have been geared toward using recommendation systems in the management of chronic diseases. Effectiveness of these systems is determined by evaluation following implementation and before deployment, using certain metrics and criteria. The purpose of this study is to ascertain whether consideration of criteria during the design of a recommendation system can increase acceptance and usefulness of the recommendation system.Design/methodology/approachUsing survey-style requirements gathering method, the specific health and technology needs of people living with chronic diseases were gathered. The result was analyzed using quantitative method. Sets of harmonized criteria and metrics were used along with requirements gathered from stakeholders to establish relationship among the criteria and the requirements. A matching matrix was used to isolate requirements for prioritization. These requirements were used in the design of a mobile app.FindingsMatching criteria against requirements highlights three possible matches, namely, exact, inferential and zero matches. In any of these matches, no requirement was discarded. This allows priority features of the system to be isolated and accorded high priority during the design. This study highlights the possibility of increasing the acceptance rate and usefulness of a recommendation system by using metrics and criteria as a guide during the design process of recommendation systems in health care. This approach was applied in the design of a mobile app called Recommendations Sharing Community for Aged and Chronically Ill People. The result has shown that with this method, it is possible to increase acceptance rate, robustness and usefulness of the product.Research limitations/implicationsInability to know the evaluation criteria beforehand, inability to do functional analysis of requirements, lack of well-defined requirements and often poor cooperation from people living with chronic diseases during requirements gathering for fear of stigmatization, confidentiality and privacy breaches are possible limitations to this study.Practical implicationsThe result has shown that with this method, it is possible to isolate more important features of the system and use them during the design process, thereby speeding up the design and increasing acceptance rate, robustness and usefulness of the system. It also helps to see in advance the likely features of the system that will enhance its usefulness and acceptance, thereby increasing the confidence of the developers in their ability to deliver a system that will meet users’ needs. As a result, developers know beforehand where to concentrate their efforts during system development to ascertain the possibility of increasing usefulness and acceptance rate of a recommendation system. In addition, it will also save time and cost.Originality/valueThis paper demonstrates originality by highlighting and testing the possibility of using evaluation criteria and metrics during the design of a recommender system with a view to increasing acceptance and enhancing usefulness. It also shows the possibility of using the metrics and criteria in system’s development process for an exercise other than evaluation.

Recently, researchers allow the analysis of the survival function of disease by the examination of the cure fraction. The extant literature discovers that the cure rate model is used for infectious but curable diseases and not infectious diseases with co-infection. Hence, a survival model that incorporates the cure rate of the management of infectious diseases with co-infection was adopted. This investigation aims to extend and develop a generalized model using the Bounded Cumulative Hazard (BCH) model, a non-mixture model for any infectious disease with a co-infection to estimate the performance of the management. The objective is to derive the appropriate probability density functions for sole infectious and co-infection disease and estimate the cure rate parameter for the two situations (sole and co-infection) using simulation data. The exponential distribution is commonly used in literature, but two-parameter Weibull distribution, a unique form of Weibull distribution, was employed in this study. The result compared using exponential distribution to accomplish the set objectives using the R package as the estimation procedure. The study concluded that the modified model, a generalized form of the cure rate model can accommodate an infectious disease with co-infection and derives the appropriate probability density function. It is also possible to extend the generalized model to other forms of distributions. This study explains the limitation of the study, the contributions, managerial implications and suggest future work.

In this paper, we study the performance of various deep reinforcement learning algorithms to enhance the energy management system of a microgrid. We propose a novel microgrid model that consists of a wind turbine generator, an energy storage system, a set of thermostatically controlled loads, a set of price-responsive loads, and a connection to the main grid. The proposed energy management system is designed to coordinate among the different flexible sources by defining the priority resources, direct demand control signals, and electricity prices. Seven deep reinforcement learning algorithms were implemented and are empirically compared in this paper. The numerical results show that the deep reinforcement learning algorithms differ widely in their ability to converge to optimal policies. By adding an experience replay and a semi-deterministic training phase to the well-known asynchronous advantage actor-critic algorithm, we achieved the highest model performance as well as convergence to near-optimal policies.

In this paper, we proposed a novel triple algorithm based on RSA (Rivest-Shamir-Adleman), AES (Advanced Encryption Standard), and TwoFish in order to further improve the security of Bluetooth that is currently using only 128-bit AES for encryption in its latest versions (Bluetooth 4.0 - 5.0). Further-more, older Bluetooth 1.0A – 3.0 + HS (High-Speed) devices use E0 stream cipher for encryption that has been shown to be weak by numerous researchers and thus it could be considered insufficient for high security purposes nowadays. In our novel approach, the triple protection of AES, RSA, and TWOFISH would enhance the level of security, which shields the data transmission in the Bluetooth. As the first step of our novel approach, we first encrypted the message by using AES with 128-bit key and then further encrypted it by using Twofish with the same 128-bit key. Finally, the 128-bit key generated in the beginning will be encrypted by using RSA with 1024-bit key to protect its over-the-air transfer. In the receiving end, the decryption process goes in reverse order compared with encryption process. We showed with experimental figures that our novel algorithm improved the security of Bluetooth encryption by eliminating all known weaknesses and thus made data exchange between Bluetooth devices secure.

In this paper, we focus on providing a narrative review of healthcare services in which artificial intelligence (AI) based services are used as part of the operations and analyze key elements to create successful AI-based services for healthcare. The benefits of AI in healthcare are measured by how AI is improving the healthcare outcomes, help caregivers in work, and reducing healthcare costs. AI market in healthcare sector have also a high market potential with 28% global compound annual growth rate. This paper will collect outcomes from multiple perspectives of healthcare sector including financial, health improvement, and care outcome as well as provide proposals and key factors for successful implementation of AI methods in healthcare. It is shown in this paper that AI implementation in healthcare can provide cost reduction and same time provide better health outcome for all.

In this chapter the appropriateness of any recommender system in healthcare, which lies in its ability to provide capabilities for meeting the challenges of modern care giving, is examined. The impacts of over two decades of research in and implementation of recommender systems in healthcare are extensively examined in two consecutive periods: first to examine empirical results and practical implementations while the second focuses on validating the earlier findings and justifying the propositions made. Although the result indicates an optimistic progress and upward trend in both the research and implementation, there are compelling reasons to invest more efforts at harmonizing evaluation criteria and metrics. In addition, in order to appropriately, adequately, and effectively meet the challenges of modern care, the rapidly evolving trends, and changing technologies, a novel solution with potential for these capabilities is proposed: a solution to provide real-time recommendations and make them available for sharing among stakeholders in real time.

published version, peerReviewed

This systematic literature review is aimed at examining empirical results and practical implementations of healthcare recommender systems. While fundamentally many of the development of recommender systems in medical and healthcare are based on theory and logic, the performance is always measured in terms of empirical results and practical implementations from evaluation of such systems. Besides, the ultimate judgment of the effectiveness of the methods and algorithms used is often based on the empirical results of recommender systems. Robustness, efficiency, speed, and accuracy are also best determined by empirical results. Extensive search was carried out in some major databases. Literature were grouped into three categories namely core, related, and relevant. The core papers were subjected to further analysis. The result shows that most work reviewed were partially evaluated and have a promising future. Moreover, a yet-to-be explored novel proposal for integration of a recommender system into smart home care is presented.

This systematic literature review is aimed at examining empirical results and practical implementations of healthcare recommender systems. While fundamentally many of the development of recommender systems in medical and healthcare are based on theory and logic, the performance is always measured in terms of empirical results and practical implementations from evaluation of such systems. Besides, the ultimate judgment of the effectiveness of the methods and algorithms used is often based on the empirical results of recommender systems. Robustness, efficiency, speed, and accuracy are also best determined by empirical results. Extensive search was carried out in some major databases. Literature were grouped into three categories namely core, related, and relevant. The core papers were subjected to further analysis. The result shows that most work reviewed were partially evaluated and have a promising future. Moreover, a yet-to-be explored novel proposal for integration of a recommender system into smart home care is presented.

Chemical imaging is an important area of Near Infrared (NIR) spectroscopy gathering a large amount of data. One of its challenges is the need for real time analysis and display. Data loss will occur when the average data acquisition rate is higher than the processing rate. This is very presumable when the spectral calculations are done in the traditional way (sequentially) on every pixel. The needed time can be shortened either by using faster processors or by distributing the computation into parallel executable parts. As the processor speed is generally limited by the available technology, the parallel computation is the only feasible improvement. Among the possible alternatives the graphics processing unit (GPU) offers a flexible, cost-effective and efficient solution. This study aims at effective parallelizing of the most important algorithms in chemical imaging using a general purpose GPU. This will be demonstrated by a comparison of execution times of sequential and parallel calculations.

In this paper, we study the concept of adaptive thresholding and present a new adaptive bottom-hat filter for the extraction of nuclei and analysis of malignancy in an image. The adaptiveness of the used structure element enables a more accurate and location-wise processing of areas in order to extract nuclei clearly for further processing such as classification of morphological features. In addition to the adaptive method we will present a color channel manipulation as a preprocessing step that adduces real benefit for the extraction phase. Some preliminary results are presented to illustrate the true potential of our method in the analysis of breast cancer. These results show that the color channel manipulation together with adaptive filtering extract the image in such a way that mainly only the nuclei is present.

A particle system, as understood in computer science, is a novel technique for modeling robots in their environment. Particle systems have traditionally been used for modeling complex dynamics of fluids and gases. In this paper, as the main contribution, we adapt our earlier work on particle systems, to compute a preliminary stress visualization for a bipedal robot walking on a soft sediment. The underlying problem of modeling rigid objects with particles is solved by introducing rack particles that enforce structural rigidness while maintaining deformability under stress. The presented approach opens many new possibilities, as it provides a computationally lightweight and unified, complementary framework for computing a stress of interacting, moving components with underspecified, non-trivial materialistic properties.

We propose two new Man-In-The-Middle (MITM) attacks on Bluetooth Secure Simple Pairing (SSP). The attacks are based on the falsification of information sent during the input/output capabilities exchange and also the fact that the security of the protocol is likely to be limited by the capabilities of the least powerful or the least secure device type. In addition, we devise countermeasures that render the attacks impractical, as well as improvements to the existing Bluetooth SSP in order to make it more secure. Moreover, we provide a comparative analysis of the existing MITM attacks on Bluetooth.

This letter presents a lossless compression method for ultraspectral sounder data. The method utilizes spectral linear prediction that exploits statistical similarities between different granules. The linear prediction with optimal granule ordering (LP-OGO) method computes linear prediction coefficients using a different granule. That approach requires one to tentatively compress all the other granules one at a time for prediction coefficient computation. The optimal ordering problem of the granules is solved by using Edmonds' algorithm. Our linear prediction with constant coefficients (LP-CC) compression method requires neither tentative compression of all the granules nor optimal ordering of the granules. We randomly select a predetermined number of granules and use that set of granules for computing constant linear prediction coefficients. Those linear prediction coefficients are used in the compression of all the other granules. The results show that the proposed method gives comparable results to the state-of-the-art method, i.e., LP-OGO, on publicly available National Aeronautics and Space Administration Atmospheric Infrared Sounder data. At the same time, the proposed method is practically applicable because it is not computationally prohibitive.

We propose an enhancement to the algorithm for lossless compression of hyperspectral images using lookup tables (LUTs). The original LUT method searched the previous band for a pixel of equal value to the pixel colocalized with the one to be predicted. The pixel in the same position as the obtained pixel in the current band is used as a predictor. LUTs were used to speed up the search. The LUT method has also been extended into a method called Locally Averaged Interband Scaling (LAIS)-LUT that uses two LUTs per band. One of the two LUT predictors that is the closest one to the LAIS estimate is chosen as the predictor for the current pixel. We propose the uniform quantization of the colocated pixels before using them for indexing the LUTs. The use of quantization reduces the size of the LUTs by an order of magnitude. The results show that the proposed method outperforms previous methods; a 3% increase in compression efficiency was observed compared to the current state-of-the-art method, LAIS-LUT.

This article presents an efficient priority pixel queue algorithm for calculating the distance transform on curved space (DTOCS), the corresponding nearest neighbor transform, and the new projection DTOCS (PDTOCS). The transforms provide tools for approximating distances and finding shortest paths on gray-level surfaces. Surface variation, or roughness, can also be measured.

In this paper, cancerous nucleus is studied from the morphological point of view with focus on the analysis of anisonucleosis, nuclear deformity, hyperchromasia, and nucleocytoplasmic ratio with CAD-based (Computer Assisted Diagnosis) solutions. Mathematical Morphology (MM), more specifically Top-Hat Transform is used together with RGB color channel manipulation to process the histopathologic image to detect and extract nuclei for further analysis. Some preliminary results are presented and they seem to be quite promising in the extraction of nuclei area. The separation of nuclei from other tissue sections is the key component in the analysis of cancerous tissues. Moreover, we will present some new ideas that will be used in our future research work.

In this paper, the current approaches and performance needs for whole slide image (WSI) analysis processing in breast cancer diagnostics are discussed. WSIs provide high resolution digital image data from the patient's diseased tissue. Digital whole slide images are typically very large and contain a high amount of information. Digitizing tissue specimen into the form of digital images allows the development and application of computational analysis algorithms. Biological tissues are complex with variance in tissue structures between healthy individuals as well as between patients with the same disease. Furthermore, the tissue preparation and digitization usually generates a lot of artifacts and more complexity, which causes classification challenges. This variance and also the large size of the images make creating an accurate and reliable automated breast cancer image analysis a challenge. In the ALMARVI project we aim at generating and implementing efficient histopathological image analysis algorithms in our breast cancer analysis scheme. This paper focuses on discussing relevant information concerning histopathological breast cancer diagnosis, and could also be considered as an introduction to the concept of WSI analysis to non-experts. Since the WSI sizes are very large (up to 40 GB with no compression) there are challenges on the computational analysis which requires computationally efficient tools and suitable approaches to relieve the problems caused by the large size of the images.

We consider geodesic distance transformations for digital images. Given a M N digital image, a distance imageis produced by evaluating local pixel distances. Distance Transformation on Curved Space (DTOCS) evaluatesshortest geodesics of a given pixel neighborhood by evaluating the height displacements between pixels. In thispaper, we propose an optimization framework for geodesic distance transformations in a pattern recognitionscheme, yielding more accurate machine learning based image analysis, exemplifying initial experiments usingcomplex breast cancer images. Furthermore, we will outline future research work, which will complete theresearch work done for this paper.Keywords: Distance transformations, feature extraction, image analysis, neural networks, optimization 1. INTRODUCTION Distance Transformations (DT) in digital images 1 is a distance mapping method for converting a binary imageinto a gray-scale image. The distance mapping is done by evaluating the distance between a foreground (feature)pixel and a background (non-feature) pixel and assigning the evaluated distance to the pixel. Distance evalu-ations done by DTs are global operations and thus for a large pixel neighborhood, the distance computationsbecomes very costly. For this reason, DTs evaluate small pixel neighborhoods at a time, still giving a reasonableapproximation to the Euclidian distance.DTs have a wide range of applications in image processing and analysis. Cuisenaire

In this article we present new lossless compression methods by combining existing methods and compare them using AVIRIS images. These methods include the Self-Organizing Map (SOM), Principal Component Analysis (PCA), and the three-dimensional Wavelet Transform combined with traditional lossless encoding methods. The two-dimensional JPEG2000 and SPIHT compression methods were applied to the eigenimages produced by the PCA. The bit allocation for the compression of eigenimages was based on the amount of information in each eigenimage. In bit rate calculation we used the exponential entropy formula, which gave better results than the original linear version. The information loss from the compression was measured by the Signal-to-Noise Ratio (SNR) and Peak-Signal-to-Noise Ratio (PSNR). To get more illustrative and practical error measures, classification of spectra was performed using unsupervised K-means clustering combined with spectral matching. Spectral matching methods include Euclidean distance, Spectral Similarity Value (SSV), and Spectral Angle Mapper (SAM). We used two test images, which both were AVIRIS images with 224 bands and 512 lines in 614 columns. The PCA in the spectral dimension combined with JPEG2000 or SPIHT in the spatial dimension was the best method in terms of the image quality and compression speed.

The distance transform on curved space (DTOCS) and its locally Euclidean modification weighted DTOCS (WDTOCS) calculate distances along gray-level surfaces. This article presents the Route DTOCS algorithm for finding and visualizing the shortest route between two points on a gray-level height map, and also introduces new distance definitions producing more accurate global distances. The algorithm is very simple to implement, and finds all optimal paths between the two points at once. The Route DTOCS is an efficient 2D approach to finding routes on a 3D surface. It also provides a more flexible solution to obstacle avoidance problems than the constrained distance transform.

A novel algorithm for the lossless compression of hyperspectral sounding data is presented. The algorithm rests upon an efficient technique for three-dimensional image band reordering. The technique is based on a correlation factor. The correlation-based band ordering gives 5% higher compression ratios than natural ordering does. On the other hand, the obtained compression ratios are within a percent of those produced by optimal ordering, but the computational time is much lower compared to the optimal ordering. The low computational complexity of the algorithm is based on the use of correlation for the band ordering. Moreover, the algorithm results in 7% to 12% improvement over fast nearest neighbor reordering scheme versions of JPEG-LS and the context-based adaptive lossless image codec algorithms.

A clustered differential pulse code modulation lossless compression method for hyperspectral images is presented. The spectra of a hyperspectral image is clustered, and an optimized predictor is calculated for each cluster. Prediction is performed using a linear predictor. After prediction, the difference between the predicted and original values is computed. The difference is entropy-coded using an adaptive entropy coder for each cluster. The achieved compression ratios presented here are compared with those of existing methods. The results show that the proposed lossless compression method for hyperspectral images outperforms previous methods.

In this paper, two new methods for edge detection in multispectral images are presented. They are based on the use of the self-organizing map (SOM) and a grayscale edge detector. With the 2-dimensional SOM the ordering of pixel vectors is obtained by applying the Peano scan, whereas this can be omitted using the 1-dimensional SOM. It is shown that using the R-ordering based methods some parts of the edges may be missed. However, they can be found using the proposed methods. Using them it is also possible to find edges in images which consist of metameric colors. Finally, it is shown that the proposed methods find the edges properly from real multispectral airplane images The size of the SOM determines the amount of found edges. If the SOM is taught using a large color vector database, the same SOM can be utilized for numerous images.

Modern smartphones and tablet PCs provide connections to the Internet via cellular networks or Wireless Local Area Networks (WLAN). The same devices also provide several wireless interfaces, such as Bluetooth for Wireless Personal Area Networks (WPANs). These interfaces create many novel possibilities for Mobile Health or Telehealth application developers to create systems providing versatile and adaptive environment for health monitoring and assisted living. In this paper, we provide a literature review based comparative analysis of current state-of-the-art mobile health applications and services. In addition, we propose and implement a mobile health platform, Smart Environment for Assisted Living (SEAL), in which smart home and mobile health functionalities are integrated into an open and adaptive Platform. The main attributes of our novel SEAL Platform are adaptivity, user-centered design, accessibility, transparency, safety, security, and privacy. Moreover, we will present some new ideas that will be used in our future research work.