A framework for medical image retrieval using machine learning and statistical similarity matching techniques with relevance feedback.

A content-based image retrieval (CBIR) framework for diverse collection of medical images of different imaging modalities, anatomic regions with different orientations and biological systems is proposed. Organization of images in such a database (DB) is well defined with predefined semantic categories; hence, it can be useful for category-specific searching. The proposed framework consists of machine learning methods for image prefiltering, similarity matching using statistical distance measures, and a relevance feedback (RF) scheme. To narrow down the semantic gap and increase the retrieval efficiency, we investigate both supervised and unsupervised learning techniques to associate low-level global image features (e.g., color, texture, and edge) in the projected PCA-based eigenspace with their high-level semantic and visual categories. Specially, we explore the use of a probabilistic multiclass support vector machine (SVM) and fuzzy c-mean (FCM) clustering for categorization and prefiltering of images to reduce the search space. A category-specific statistical similarity matching is proposed in a finer level on the prefiltered images. To incorporate a better perception subjectivity, an RF mechanism is also added to update the query parameters dynamically and adjust the proposed matching functions. Experiments are based on a ground-truth DB consisting of 5000 diverse medical images of 20 predefined categories. Analysis of results based on cross-validation (CV) accuracy and precision-recall for image categorization and retrieval is reported. It demonstrates the improvement, effectiveness, and efficiency achieved by the proposed framework.     

Real-Time Testing of WPT-Based Protection of Three-Phase VS PWM Inverter-Fed Motors

This paper presents an experimental testing of a wavelet packet transform (WPT)-based technique for protecting and controlling three-phase voltage-source (VS) pulse-width modulated (PWM) inverters under extreme load conditions. The proposed technique is based on a two-level multiresolution analysis (MRA) applied on output currents of a 3Phi VS PWM inverter. The MRA aims to extract certain signatures, which are the second level highest frequency subband coefficients. These coefficients take a nonzero value for any signal that is short-duration, nonperiodic and nonstationary with impulse-super imposed high frequencies. Such a signal perfectly matches a current arising from any typical fault occurring in either the inverter legs or on the load side. The proposed technique is realized using a -code hosted and executed by a dSPACE ds1102 controller board. The tested inverter supplies loads that include a 3Phi Y-connected R - L load, a 3Phi squirrel-cage induction motor and a 3Phi synchronous reluctance motor. The extreme load conditions include inverter open-leg, motor starting currents and different faults in motor stator windings. The experimental test results show accurate, fast, and effective response to all disturbances including fault currents by the proposed WPT-based technique.     

On pole distribution in digital filters

Due to the discrete nature of the coefficients the poles and zeros of a digital filter transfer function cannot, in general, be realized exactly. This paper introduces a new concept ‘pole sparsity factor’ as a measure of the maximum possible pole deviation after tuning, for the case of second order circuits. A new second order circuit that exhibits lower pole sparsity factor compared to presently available circuits with the same number of multipliers is presented.     

Modelling Energy Dissipation Over Stepped-gabion Weirs by Artificial Intelligence

The hydraulics of energy dissipation over stepped-gabion weirs is investigated by carrying out a series of laboratory experiments, building models to explain the experimental data, and testing their robustness by using the data reported by other researchers. The experiments comprise: six different stepped-gabion weirs tested in a horizontal laboratory flume, a wide range of discharge values, two weir slopes (V:H): 1:1 and 1:2, and gabion filling material gravel size (porosity equal to 38 %, 40 % and 42 %). These experimental setups were selected to ensure the development of both the nappe and skimming flow regimes within the measured dataset. The models developed for computing energy dissipation over stepped-gabion weirs comprise: multiple regression equations based on dimensional analysis theory, Artificial Neural Network (ANN) and Gene Expression Programming (GEP). The analysis shows that the measured data capture both flow regimes and the transition in between them and above all, and by using all of the data, it may be possible to identify the range of each regime. Energy dissipation modelled by the ANN formulation is successful and may be recommended for reliable estimates but those by GEP and regression analysis can still serve for rough-and-ready estimates in engineering applications.     

Comparing the performance of multispectral vegetation indices and machine-learning algorithms for remote estimation of chlorophyll content: a case study in the Sundarbans mangrove forest

Optical vegetation indices (VIs) have been used to retrieve and assess biophysical variables from satellite reflectance data. These indices, however, also are sensitive to a number of confounding factors, such as canopy geometry, soil optical properties, and solar position. This suggests that VIs should be used cautiously for biophysical parameter estimation. Among biophysical variables, chlorophyll content is of particular importance as an indicator of photosynthetic activity. The goal of this study is to investigate the performance of multispectral optical VIs for chlorophyll content estimation in the world’s largest mangrove forest, the Sundarbans, and to compare these with machine-learning algorithms (MLAs). To this end, we have investigated the performance of 15 multispectral VIs and six state-of-the-art MLAs that are widely used for adaptive data fitting. The MLAs are Artificial Neural Networks (ANNs), Genetic Algorithm (GA), Gaussian Processes for Machine Learning (GPML), Kernel Ridge Regression (KRR), Locally Weighted Polynomials (LWP), and Multivariate Adaptive Regression Splines (MARS). We use an in situ data set of reflectance and chlorophyll measurements to develop and validate our models. Each MLA was evaluated 500 times with random partitions of training and validation data. Results showed that the weight optimization and term selection used within GA produce the most reliable chlorophyll content estimation. However, green normalized difference VI (GNDVI) is a simple and computationally efficient VI that produces results that are nearly as accurate as GA in terms of model fit and performance. Results also show that all methods except ANNs and MARS produce a quasi-linear relationship between spectral reflectance and chlorophyll content. Statistical transformations of GNDVI and chlorophyll content have the capability of further reducing model error.     

Modelling and prediction of malaria vector distribution in Bangladesh from remote-sensing data

Epidemic malaria cases and satellite-based vegetation health (VH) indices were investigated to be used as predictors of malaria vector activities in Bangladesh. The VH indices were derived from radiances, measured by the Advanced Very High Resolution Radiometer (AVHRR) on National Oceanic and Atmospheric Administration (NOAA) afternoon polar orbiting satellites. Two indices characterizing moisture and thermal conditions were investigated using correlation and regression analysis applied to the number of malaria cases recorded in the entire Bangladesh region and three administrative divisions (Chittagong, Sylhet and Dhaka) during 1992–2001. It is shown that during the cooler months (November to March), when mosquitoes are less active, the correlation between number of malaria cases and two investigated indices was near zero. From April, when the mosquito activity season starts, the correlation increased, reaching a maximum value of 0.5–0.8 by the middle of the high season (June to July), reducing thereafter to zero by the beginning of the cool season in November. Following these results, regressional equations for the number of malaria cases as a function of VH indices were built and tested independently. They showed that, in the main malaria administrative division (Chittagong) and the entire Bangladesh region, the regressional equations can be used for early prediction of malaria development.     

Expert system classification of urban land use/cover for Delhi,India

This study presents the results of classifying land use/land cover for Delhi, India using an expert system approach. For this study Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data of 22 September 2003 were used. The research goals of this project are two‐fold. In one respect, the research goal is to report on the extent covered by urbanization using the classified image. Thirteen different land‐cover categories were identified with an 85.55% overall classification accuracy based on 256 random points for validation and 50 on the ground observations. Secondly, we report on our efforts to duplicate an expert system model previously developed for Phoenix Arizona as a generalized approach for urban land use classification. Results suggest that while some of the methodology could be duplicated, there are local factors (e.g. data availability and specific land features) that required the approach to be modified.     

Fabrication of different geometry cutting tools and their effect on the vertical micro-grinding of BK7 glass

With the demand for microstructures of not only with diversified shape but also of reduced dimension on glass, fabrication of polycrystalline diamond (PCD) tool/microelectrodes with different shape has become important. However, to date, fabrication of different shapes in single setup is not possible and also needs special indexing attachment. To solve this problem, in this study, a specially designed block containing three v-slots of 60°, 90°, and 120° has been designed and fabricated using wire cut. Thereafter with the help of block electro-discharge machining method and using this specially designed block, different shapes of microelectrodes with symmetrical and non-symmetrical section has been fabricated. This study also investigates the feasibility of using these different geometry PCD tool for micro-grinding of BK7 glass. In this context, a relative study on the micro-grinding performance of four different geometry tools (circular, D-shaped, triangular, and square) has been carried out. It has been observed that among the different shaped tools, D-shaped tool experienced lowest cutting force along x- and y-axes where as triangular tool faced lowest force along z-axis, and highest cutting forces were found to be experienced by square tool. Average and maximum roughness of machined surface was found to be improved from circular to others tool except triangular one. But, it was also observed that side surface started to deteriorate from circular to other tool due to edge wear. In case of tool wear, square and triangular tool experienced more wear than circular and D-shaped tool due to their frequent edge blunting or rounding effect. Finally, among four different geometry tools, D-shaped tool was considered to provide better performance in terms of the achieved surface finish, tool wear, and cutting force analysis.     

Performance evaluation of endrills

This paper evaluates the performance of a relatively new type of drill called an endrill which is a cross between a drill and an endmill. Investigations into the effects of its cutting conditions on the drilling forces, surface finish, drill wear and hole oversize were carried out. It was found that endrills produced better quality holes than conventional twist drills, better surface finish and less oversize of the holes. Hence, with proper feed, speed and flow rate of the pressurized flushing coolant, a good finish of about R_a = 1 μm can be attained without reaming. Thus, the productivity of finished holes can be remarkably improved. Compared to twist drills, lower torque and thrust were observed which yielded improved tool life and reduced power consumption. No “walking phenomenon” was observed when this kind of drill was used and the amount of hole oversize was found to average about 0.7% of the drill diameter as compared to 1.6% when twist drills were used. Finally, general equations for the drill torque and thrust were derived from the experimental results.