Stefan Mocanu

Benign and malignant breast tumors: Diagnosis using fractal measures

The work presents two measures of complexity, fractal dimension and lacunarity, in order to raise the precision in breast cancer diagnosis. A set of 40 cases of mammograms from patients corresponding to both benign (24 images) and malignant tumors (16 images) were analyzed. To improve the diagnostic process we proposed a method that combines the two fractal characteristics. For the processing of mammograms it was used two software programs, one for computing average fractal dimensions from the image contour, proposed by the authors, and the other (FracLac) to compute the average lacunarity on the binary image. In this way, classification rate increased from 90% (when using fractal dimension) to 100%. Finally, we proposed a framework for assisted diagnosis from the mentioned set of mammographic images.

Audience indicators for geospatial marketing using traffic data from cellular networks

This paper proposes a modern approach for the methodology that supports the measure the audience for outdoor poster advertising. The classic approach of conducting complex, limited and expensive studies about human daily routes, related just to a limited type of advertising poster may be enriched using the data that resides in the monitoring infrastructure of the cellular telecom operators over digital maps. Well segmented, this kind of data may reveal important features of the population activity at a large scale. Finally, the fine tuning options in analyzing this kind of data are discussed for future improvements.

Towards Practical Integration of WSN in Cloud Dedicated to Smart Environments

In this paper, a method of interaction with Wireless Sensors Networks (WSN) through cloud, called Wireless Sensor Network over Cloud (WSN-C), is proposed. The physical sensors are not considered stand-alone sensors but they can be monitored, controlled and updated (programmed) via cloud services. The Cloud can communicate with the sensors through the WSN access point, or Sensor Gateway. Thus, every sensor inside the network can be accessed, monitored and controlled from the Cloud. In this way, the user not only has access to the raw data from physical sensors, but also can create virtual sensors and can program and interact directly with the physical sensor, remotely via cloud services. In addition the sensing layer represented by the physical sensors, an actuating layer (based on mechanical, electrical or hydraulic actuators) is included. The Cloud can communicate with the actuators through the Actuator Gateway. There are many applications in which such architecture is needed, especially for monitoring physical sensors, allowing multiple users access to the data, providing higher security, scalability and a way of processing data and send back commands to the local node.

Differential Equations Systems versus Scale Free Networks in Sepsis Modeling

The paper proposes a comparison between models that relieves the characteristics of the patient evolution in sepsis as the result of correct or corrupted information transfer as the cell level. Specifically, two different models for the cellular interactions are compared - one based on a three variables system of differential equations, the other on a cellular interaction scale free network. The advantages and limits of each model in understanding the structure of cellular signaling networks involved in sepsis phenomena are finally discussed

Process simulation platform for virtual manufacturing systems evaluation

Abstract The paper presents the design of a cloud simulation platform and the associated services that will provide the computing resources and services for hybrid simulation of Virtual Manufacturing Systems. At the core of the research is the design of a framework defined as Virtual Development Environment (VDE). Associated to VDE is a set of block functions and algorithms defined as software assets, stored in a large repository implemented within an on-line cloud supported library. The platform offers flexibility concerning the continuous testing and updates regarding the algorithms in order to improve the existing ones.

Secure Communication Method Based on Encryption and Steganography

Encryption is a traditional method of scrambling data beyond recognition. For years, encryption has been widely used in various areas and domains where secrecy and confidentiality were required. However, there are certain encryption techniques that are not allowed in some countries. In these cases, steganography may come as a solution for hiding data using an apparently inoffensive and innocent carrier. The purpose of steganography is to deliver secret information from a sender to a receiver by scrambling the communication channel not the information itself. This way, it is very unlikely for someone to identify and extract the secret message without knowing exactly how it was embedded into the carrier. Moreover, this task is more difficult since most steganographic methods rely on file or packet header redundancy, lossy multimedia algorithms or insignificant content alteration. All of these are very hard to detect. In this paper, a combination of encryption with image based steganography is presented in order to offer similar robustness as offered by advanced encryption algorithms such as RSA, AES or DES but with lower resource needs. The most original contribution of this paper consists in how a secret message is embedded into a image carrier generating 3 different but very similar images. These images are sent to the destination using different channels and then used to recover the original (secret) message.

Optimizing OOH Media Plans Using Weighted Set Cover Algorithms

Media Plan Optimizations (MPOs) are frequently employed in the advertising industry. Given a variety of advertising spaces, the goal is to have an impact as high as possible while minimizing costs. Taking into account that each advertising space has a given cost and has its own viewers, and two spaces can share common viewers, this translates into a weighted set cover problem. Problems of this type are NP-complete, so there are no known polynomial-time exact algorithms for finding the optimal solution. An alternative is to use a Greedy algorithm, which offers a good approximation in a reasonable amount of time. In this paper, we study the mapping of the Media Plan Optimization problem to the Weighted Set Cover problem. We analyze the performance of the Backtracking and Greedy algorithms in respect to computational time and solution quality, under a wide range of OOH (Out-Of-Home) MPO scenarios. We evaluate the possibility of using these algorithms in a high availability, client-server Media Planning solution based on web services and find the computational performance and accuracy of the Greedy Algorithm to be adequate for the task.

Complex networks to model the economic globalization process

Abstract The paper analyses the effects of the process of globalization not only in economic terms, but rather the profound change in the structural properties of the socio-economic networks towards more complexity (more connectivity, more interactions, faster changes and greater uncertainties). We apply network metrics to international trade data and show a clear trend towards a rise of complexity in the patterns of global economic exchanges. We suggest that the observed increase in scale and connectivity of the common economic networks induces greater uncertainties and risks of large-scale instabilities and propose some efficient means to manage network complexity based on efficient models.

Visual inspection system for automated sorting of seeds

Abstract This paper discusses a method to evaluate the potential of texture analysis in the automated inspection and sorting of seeds. Two special procedures for image processing considered the most suitable for this aim are presented: one that allows to compute the energy of the sub bands of a Wavelet Transform, the other that determines the local and connected fractal dimension of an image. Then an interactive decision support system that uses for the classification of images the cumulative feature vector including the wavelet coefficients and the fractal dimension is described, together the procedure to take decisions on the image relevance.