Ioannis N. Dimou

Survivin regulation by HER2 through NF-κB and c-myc in irradiated breast cancer cells

Radiotherapy is an important treatment modality against cancer resulting in apoptosis and inhibition of cell growth. Survivin is an important cancer biomarker conferring to tumour cells increased survival potential by inhibiting apoptosis. In the present study, we investigated the implication of breast cancer cells features, as hormone receptors and p53 status, in the radio‐resistance of breast cancer cells and in the regulation of survivin’s expression by nuclear factor (NF)‐κB and c‐myc. Six breast cancer cell lines Michigan Cancer Foundation (MCF‐7), MCF‐7/Human Epidermal Growth Factor Receptor (HER)2, M. D. Anderson – Metastatic Breast (MDA‐MB‐231), SK‐BR‐3, BT‐474 and Human Breast Lactating (HBL‐100) were irradiated and cell viability as well as cell cycle distribution were evaluated by 3‐(4,5‐Dimethylthiazol‐2‐yl)‐2,5‐diphenyltetrazolium bromide (MTT) assay and flow cytometry, respectively. Survivin mRNA and protein levels were evaluated by real time PCR and Western blot analysis. Survivin and HER2 gene knockdown was performed with siRNA technology and investigation of transcription factors binding to survivin and c‐myc gene promoters was assessed by chromatin immunoprecipitation. Student’s t‐test and F‐statistics were used for statistical evaluation. Our results demonstrated that only HER2+ breast cancer cells up‐regulated survivin upon irradiation, whereas HER2 knockdown in HER2+ cells led to survivin’s down‐regulation. Survivin and especially HER2 knockdown abolished the observed G2/M cell cycle checkpoint and reduced the radio‐resistance of HER2 overexpressing breast cancer cells. Additionally, HER2 was found to regulate survivin’s expression through NF‐κB and c‐myc transcription factors. This study revealed the significance of HER2 in the radio‐resistance of HER2+ breast cancer cells through induction of transcription factors NF‐κB and c‐myc, leading to activation of survivin, a downstream target oncogene preventing apoptosis.

Combination of multiple classifiers for post-placement quality inspection of components: A comparative study

Current trends in the electronics industry are towards miniaturization of components, denser packing of printed-circuit boards and highly automated assembly lines. The technology of Surface Mounted Devices (SMD) facilitates this trend, thus explaining the substantial increase in the use of its various versions. Nevertheless, dense packaging requires increased accuracy in the placement and efficient inspection of components in order to ensure high reliability in manufacturing. This paper presents fusion methods of multiple classifiers for improving the classification of individual components in terms of positioning accuracy through computer vision inspection. Multiple classifier combination is a technique that combines the decisions of different classifiers as to reduce the variance of estimation errors and improve the overall classification accuracy. Combining the power of the primary classifiers through multi-modular architectures improves the classification results and contributes to the robustness of the overall inspection system.

Brain lesion classification using 3T MRS spectra and paired SVM kernels

Abstract The increased power and resolution capabilities of 3T Magnetic Resonance (MR) scanners have extended the reach of Magnetic Resonance Spectroscopy as a non-invasive diagnostic tool. Practical sensor calibration issues, magnetic field homogeneity effects and measurement noise introduce distortion into the obtained spectra. Therefore, a combination of robust preprocessing models and nonlinear pattern analysis algorithms is needed in order to evaluate and map the underlying relations of the measured metabolites. The aim of this work is threefold. Firstly we propose the use of a paired support vector machine kernel utilizing metabolic data from both affected and normal voxels in the patients brain for lesion classification problem. Secondly we quantify the performance of an optimal reduced feature set based on targeted CSI-144 scans in order to further reduce the data volume required for a reliable computed aided diagnosis. Thirdly we expand our previous formulation to full multiclass classification. The long term aim remains to provide the human expert with an easily interpretable system to assist clinicians with the time, volume and accuracy demanding diagnostic process.

Classification of pathological human brain lesions using Magnetic Resonance Spectroscopy at 3T

Magnetic Resonance Spectroscopy is a powerful non-invasive diagnostic tool that is used in conjunction with MRI techniques to provide identification and quantification of biologically important compounds in soft tissue. However in diagnostic applications the underlying relations of the measured metabolites’ values that the clinician has to take into account are too complex to be coded into simple decision rules. Moreover magnetic field homogeneity effects, measurement noise and sensor calibration issues induce distortion into the obtained spectra. In this work we focus on utilizing a state of the art support vector machine classification system to undertake the task of brain tumor classification. We aim at providing the human expert with easily interpretable probabilistic metrics to assist in the time, volume and accuracy demanding diagnostic process.

A Hopfield neural network based ATM routing method for the IRIDIUM system

An efficient model for ATM routing in the IRIDIUM system is presented in this paper. The model tries to be as realistic as possible. A proper energy function is constructed from the constraints of the ATM routing concept which are clearly indicated. The combinatorial nature of the problem is clearly shown and a properly modified Hopfield neural network can be easily applied and give adequate solutions.

A Statistical Diagnostic Decision Support Tool Using Magnetic Resonance Spectroscopy Data

The aim of this study was to develop a practical postprocessing statistical tool for spectroscopic data analysis to successively create an effective noninvasive tool for spectra evaluation and cerebral disease classification. Spectroscopic data were obtained from a total of 112 patients suffering from several brain lesions. The evaluation was based on histological diagnosis, and/or radiological diagnoses and/or medical physicists’ observation. First, calculation of metabolite ratio (NAA/Cr, Cho/Cr, mI/Cr, LL/Cr) means was conducted for each pathological case, and results were compared with the corresponding published data. A Matlab-based algorithm called FA.S.M.A (Fast Spectroscopic Multiple Analysis) with a Graphical User Interface (GUI) was developed, performing nearest mean classification. It is a fast and user-friendly radiological tool which provides fundamental functionality in estimating mean metabolite ratios values during spectroscopy examination. The user can insert the metabolite ratios and obtain the most probable diagnostic class and the corresponding mean spectrum based on published prior knowledge. In future, FA.S.M.A will be extended to enrich more advanced Pattern Recognition techniques and additional machine learning (ML) methods will be implemented in order to provide a more accurate mapping of the input data to facilitate brain tumor classification according to histological subtype. From a clinical point of view, FA.S.M.A will be extended to incorporate quantitative data from other advanced MR-based techniques such as DWI, DTI, and perfusion measurements not only for supporting primary diagnosis of tumor type but also for determining the extent of glioma infiltration with a high degree of spatial resolution.

A generalized-space expansion of Support Vector Machines for diagnostic systems

Support Vector Machines (SVMs) are by now an established tool used in state of the art applications in the biomedical domain. Their prevalence has unveiled both a very effective generalization capability and the inherent positive definiteness constraints in kernel selection. In this work we apply a series of composite kernel extensions stemming from nonlinear second-level kernels to standard diagnostic problems. Our aim is twofold. Firstly, to create a formulation that can accept arbitrary non-positive definite feature kernels and secondly, to allow for nonlinear second-level kernels as part of this scheme.

Error bounds of decision templates and support vector machines in decision fusion

The need for accurate, robust, optimised classification systems has been driving information fusion methodology towards a state of early maturity throughout the last decade. Among its shortcomings we identify the lack of statistical foundation in many ad-hoc fusion methods and the lack of strong non-linear combiners with the capacity to partition complex decision spaces. In this work, we draw parallels between the well known decision templates (DT) fusion method and the nearest mean distance classifier in order to extract a useful formulation for the overall expected classification error. Additionally we evaluate DTs against a support vector machine (SVM) discriminant hyper-classifier, using two benchmark biomedical datasets. Beyond measuring performance statistics, we advocate the theoretical advantages of support vectors as multiple attractor points in a hyper-classifiers feature space.

Security aspects of inter-satellite links using neural networks

This paper presents a neural network model for routing in the space segment of a Satellite Personal Communication System. At first, a proper energy function is constructed from the constraints of the system, which are clearly indicated. The combinatorial nature of the problem is clearly shown and hence a properly modified Hopfield neural network can be easily applied to provide increased survivability and thus improve security.