Search

Diagnosing Alzheimer’s Disease (AD) in older people using magnetic resonance imaging (MRI) is quite hard since it requires the extraction of highly discriminative feature representation from similar brain patterns and pixel intensities. However, deep learning techniques possess the capability of extracting relevant representations from data. In this work, we designed a novel spiking deep convolutional neural network-based pipeline to classify AD using MRI scans. We considered three MRI scan groups (patients with AD dementia, Mild Cognitive Impairment (MCI), and healthy controls (NC)). We developed a three-binary classification task (AD vs. NC, AD vs. MCI, and NC vs. MCI) for the AD classification tasks. Specifically, an unsupervised convolutional Spiking Neural Networks (SNN) is pre-trained on the MRI scans. Finally, a supervised deep Convolution Neural Network (CNN) is trained on the output of the SNN for the classification tasks. Experiments are performed using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, and promising results are obtained for the AD classification tasks. We present our proposed model results for both the unsupervised spike pre-training technique and the case where the pre-training technique was not considered, thus serving as a baseline. The accuracy of the proposed model with spike pre-training techniques for the three-binary classification are 90.15%, 87.30%, and 83.90%, respectively, and the accuracy of the model without the spike are 86.90%, 83.25%, and 76.70%, respectively, with a noticeable increase in accuracy and thus, reveals the effectiveness of the proposed method. We also evaluated the robustness of our proposed approach by running experiment on six baseline methods using our preprocessed MRI scans. Our model outperformed almost all the comparable methods due to the robust discriminative capability of the SNN in extracting relevant AD features for the AD classification task.

In this paper, we present an extension on a hybrid-based deduplication technique in entity reconciliation (ER), by proposing an algorithm that builds clusters upon receiving a pre-specified K number of clusters, and second developing a crowd-based procedure for refining the results of the clusters produced after the clustering generation phases. With the clusters refined, we aim to minimize the cost metric Λ'(R) of the solitary and compound cluster generation algorithms, to achieve an improved and efficient deduplication method, to have an increase in accuracy in identifying duplicate records, and finally, further reduce the crowdsourcing overheads incurred. In this paper, in the experiments, we made use of three datasets commonly known to hybrid-based deduplication such as paper, product, and restaurant. The performance results and evaluations demonstrate clear superiority to the methods compared with our work offering low-crowdsourcing cost and high accuracy of deduplication, as well as better deduplication efficiency due to the clusters being refined.

Deep learning architecture has shown remarkable performance in machine learning and AI applications. However, training a spiking Deep Convolutional Neural Network (DCNN) while incorporating traditional CNN properties remains an open problem for researchers. This paper explores a novel spiking DCNN consisting of a convolutional/pooling layer followed by a fully connected SNN trained in a greedy layer-wise manner. The feature extraction of images is done by the spiking DCNN component of the proposed architecture. And in achieving the feature extraction, we leveraged on the SAILnet to train the original MNIST data. To serve as input to the convolution layer, we process the raw MNIST data with bilateral filter to get the filtered image. The convolution kernel trained in the previous step is used to calculate the filtered image’s feature map, and carry out the maximum pooling operation on the characteristic map. We use BP-STDP to train the fully connected SNN for prediction. To avoid over fitting and to further improve the convergence speed of the network, a dynamic dropout is added when the accuracy of the training sets reaches 97% to prevent co-adaptation of neurons. In addition, the learning rate is automatically adjusted in training, which ensures an effective way to speed up training and slow down the rising speed of the training accuracy at each epoch. Our model is evaluated on the MNIST digit and Cactus3 shape datasets, with the recognition performance on test datasets being 96.16% and 97.92% respectively. The level of performance shows that our model is capable of extracting independent and prominent features in images using spikes.

The Bat Algorithm (BA) is a meta-heuristic algorithm based on echolocation behavior of microbats. The authors propose BA based Spiking Neural Network (SNN) model, where the advantages of BA and efficiency of SNN are exploited for classification tasks using some benchmark datasets. The advantages of the BA have been well exploited in the Artificial Neural Networks (ANN) domain particularly with the adjustment of weights. We therefore, leveraged on the BA as a learning strategy to train an SNN using the Leaky Integrate and Fire (LIF) and Izhikevich models to solve non-linear pattern classification tasks. In order to successfully discriminate between the various classes, the models are trained to fire at the same or similar firing rate for inputs from the same class, and inputs patterns from different classes to also spike or fire at different rate. To justify how efficient and how powerful the proposed model is, only one neuron is used. Finally, the model is tested on different non-linear pattern recognition tasks and comparison is made between our model and other similar existing models and our proposed model outperformed some of the state-of-the-art-models. To the best of our knowledge, this is the first work to implement BA in SNN.

Cataracts are the leading cause of blindness worldwide. Current methods for discriminating cataractous lenses from healthy lenses of Sprague-Dawley rats during preclinical studies are based on either histopathological or clinical assessments which are weakened by subjectivity. In this work, both cataractous and healthy lens tissues of Sprague-Dawley rats were studied using multispectral imaging technique in combination with multivariate analysis. Multispectral images were captured in transmission, reflection and scattering modes. In all, five spectral bands were found to be markers for discriminating cataractous lenses from healthy lenses; 470 nm and 625 nm discriminated in reflection mode whereas 435 nm, 590 nm and 700 nm discriminated in transmission mode. With Fisher’s Linear discriminant analysis, the midpoints for classifying cataractous from healthy lenses were found to be 14.718 × 10−14 and 3.2374 × 10−14 for the two spectra bands in the reflection mode and the three spectral bands in the transmission mode respectively. Images in scattering mode did not show significant discrimination. These spectral bands in reflection and transmission modes may offer potential diagnostic markers for discriminating cataractous lenses from healthy lenses thereby promising multispectral imaging applications for characterizing cataractous and healthy lenses.

Laser-induced fluorescence (LIF) combined with multivariate techniques has been used in identifying antimalarial herbal plants (AMHPs) based on their geographical origin. The AMHP samples were collected from four geographical origins (Abrafo, Jukwa, Nfuom, and Akotokyere) in the Cape Coast Metropolis, Ghana. LIF spectra data were recorded from the AMHP samples. Utilizing multivariate techniques, a training set for the first two principal components of the AMHP spectra data was modeled through the use of K-nearest neighbor (KNN), support vector nachine (SVM), and linear discriminant analysis (LDA) methods. The SVM and KNN methods performed best with 100% success for the prediction data, while the LDA had a 99% success rate. The KNN and SVM methods are recommended for the identification of AMHPs based on their geographical origins. Deconvoluted peaks from the LIF spectra of all the AMHP samples revealed compounds such as quercetin and berberine as being present in all the AMHP samples.

Laser-induced autofluorescence (LIAF), combined with multivariate techniques, has been used to discriminate a cataractous lens from healthy lens tissues. In this study, 405 nm and 445 nm were used as excitation sources to induce the autofluorescence. Results show higher autofluorescence intensity in cataractous lens tissues than in healthy ones. Cataractous lens tissues show a red shift of 0.9 nm and 1.2 nm at 405 nm and 445 nm excitations, respectively. Using principal component analysis (PCA), three principal components (PCs) gave more than 99% variability for both 405 nm and 445 nm excitation sources. Based on the three PCs, Fisher’s linear discriminant model was developed. An accuracy of 100% was obtained in classifying the lens tissues using Fisher’s linear discriminant analysis (FLDA). The LIAF technique assisted by PCA and FLDA may be used for objective discrimination of cataractous lens from healthy lens tissues of Sprague–Dawley rats.

In real world, entities may occur several times in a database. These duplicates may have varying keys and/or include errors that make deduplication a difficult task. Deduplication cannot be solved accurately using either machine-based or crowdsourcing techniques only. Crowdsourcing were used to resolve the shortcomings of machine-based approaches. Compared to machines, the crowd provided relatively accurate results, but with a slow execution time and very expensive too. A hybrid technique for data deduplication using a Euclidean distance and a chromatic correlation clustering algorithm was presented. The technique aimed at: reducing the crowdsourcing cost, reducing the time the crowd use in deduplication and finally providing higher accuracy in data deduplication. In the experiments, the proposed algorithm was compared with some existing techniques and outperformed some, offering an utmost deduplication accuracy efficiency and also incurring low crowdsourcing cost.

Plasmodium falciparum (P. falciparum) malarial degree of infection, termed as parasite density (PD), estimation is vital for point-of-care diagnosis and treatment of the disease. In this work, we present application of optical techniques: optical absorption and multispectral imaging for P. falciparum malarial byproduct (hemozoin) detection in human‐infected blood samples to estimate PD. The blood samples were collected from volunteers who were tested positive for P. falciparum infections (i-blood), and after treatment, another set of blood samples (u-blood) were also taken. The i-blood samples were grouped based on PD (+, ++, +++, and ++++). Optical densities (ODs) of u-blood samples and i-blood samples at blood absorption bands of 405 nm, 541 nm, and 577 nm showed different optical absorption characteristics. Empirical computation of ratio of the ODs for the blood absorption bands revealed reduction in the ODs with increasing PD. Multispectral images containing uninfected red blood cells (u-RBCs) and P. falciparum‐infected red blood cells (i-RBCs) on unstained blood smear slides exhibited spectrally determined decrease in both reflected and scattered pixel intensities and increase in transmitted pixel intensities with increasing PD. We further propose a linear classification model based on Fisher’s approach using reflected, scattered, and transmitted pixel intensities for easy and inexpensive estimation of PD as an alternative to manual estimation of PD, currently, the widely used technique. Application of the optical techniques and the proposed linear classification model are therefore recommended for improved malaria diagnosis and therapy.

In real world, databases often have several records representing the same entity and these duplicates have no common key, thus making deduplication difficult. Machine-based and crowdsourcing techniques were disjointly used in improving quality in data deduplication. Crowdsourcing were used for solving tasks that the machine-based algorithms were not good at. Though, the crowds, compared with machines, provided relatively more accurate results, both platforms were slow in execution and hence expensive to implement. In this paper, a hybrid human-machine system was proposed where machines were firstly used on the data set before the humans were further used to identify potential duplicates. We performed experiments using three benchmark datasets; paper, restaurant and product datasets. Our algorithm was compared with some existing techniques and our approach outperformed some methods by achieving a high accuracy of deduplication and good deduplication efficiency while incurring low crowdsourcing costs.

Entity resolution (ER) classifies records that refer to the same real-world entity and is fundamental to data cleaning. Identifying approximate but not exact duplicates in database records is a vital task. These duplicates may refer to the same real-world entity due to; 1. data entry errors, 2. differences in the detailed schemas of records from multiple databases, 3. unstandardized abbreviations, among several reasons. Machine-based techniques have been improving in quality, but still are a long way from being perfect. On the other hand, crowdsourcing platforms are widely accepted as a means for resolving tasks that machine-based approaches are not good at. Though they also offer relatively more accurate results, they are expensive and are slow in bringing human perception into the process.

The optical properties of the transparent wings of the Dragonfly Aeshna cyanea were studied using a comprehensive set of optical measurements, experimental analysis and theoretical modeling which involves the use of a high level programming language to simulate the optical effects seen. With these, the relative refractive index of the Odonatan wing, the pruinosity associated with the microstructure and the chemical composition of the wings were studied. The Nystrom matrix techniques were applied to solve the surface currents JZ and HZ of the scattered fields for different incident angles from grazing and used to explain the pruinosity associated with the wings microstructure. The wing was found to be an Electro-Optic Material (EOM) associated with a number of Nonlinear Optical (NLO) responses having high frequency absorption for extreme UV and also, leaky multi-channeling wave guide.

Violet diode laser-induced chlorophyll fluorescence was used in agronomical assessment (disease severity and average yield per plant). Because cassava (Manihot esculenta Crantz) is of economic importance, improved cultivars with various levels of affinity for cassava mosaic disease were investigated. Fluorescence data correlated with cassava mosaic disease severity levels and with the average yield per plant.

Floodplain wetlands influence the timing and magnitude of stream responses to rainfall. In managing and sustaining the level of water resource usage in any river catchment as well as when modelling hydrological processes, it is essential that the role of floodplain wetlands in stream flows is recognised and understood. Existing studies on hydrology within the Volta River basin have not adequately represented the variability of wetland hydrological processes and their contribution to the sustenance of river flow. In order to quantify the extent of floodwater storage within riparian wetlands and their contribution to subsequent river discharges, a series of complementary studies were conducted by utilising stable isotopes, physical monitoring of groundwater levels and numerical modelling. The water samples were collected near Pwalugu on the White Volta River and at three wetland sites adjacent to the river using the grab sampling technique. These were analysed for (18)O and (2)H. The analysis provided an estimate of the contribution of pre-event water to overall stream flow. In addition, the variation in the isotopic composition in the river and wetland water samples, respectively, revealed the pattern of flow and exchange of water between the wetlands and the main river system.

An opto-digital imaging system for counting and analysing tracks on a LR-115 detector is described. One batch of LR-115 track detectors was irradiated with Am-241 for a determined period and distance for linearity test and another batch was exposed to radon gas. The laser-illuminated etched track detector area was imaged, digitized and analysed by the system. The tracks that were counted on the opto-digital system with the aid of media cybernetics software as well as spark gap counter showed comparable track density results ranging between 1500 and 2750 tracks cm?2 and 65 tracks cm?2 in the two different batch detector samples with 0.5% and 1% track counts, respectively. Track sizes of the incident alpha particles from the radon gas on the LR-115 detector demonstrating different track energies are statistically and graphically represented. The opto-digital imaging system counts and measures other track parameters at an average process time of 3?5 s.

DNA origami technology provides a versatile approach for the chemical assembly of gold nanostructures. In this study the bottom-up approach of self-assembly using DNA in the origami process has been successfully applied to arrange five AuNPs asymmetrically. The DNA origami templates were modified to have binding sites that were extended with sticky ends to facilitate the attachment of the AuNPs. With the help of thiol chemistry, the AuNPs which were covered with DNA complementary to the sticky ends introduced on the DNA origami surfaces, we were able to attach the nanoparticles to the designed sites. It was realized that there were slight differences in the designed distances and the determined ones which were accounted for potentially by the deposition of the structures on the grids for imaging. The structures were characterized with gel electrophoresis and TEM. This asymmetric arrangement has the potential of exhibiting plasmonic behavior and circular dichroism when light is incident on the structure.

Multi-spectral imaging (MSI) has made diagnosis of microscopic samples considerably easier and information abound. Most MSI systems use continuum light sources and filters for imaging purposes. However, these light sources and filters are relatively expensive, unstable due to extreme pressure and temperature and associated with prolong acquisition time. In this work, we present a metallurgical microscope retrofitted with light-emitting diodes (LEDs) as illumination sources for MSI microscopy. This multispectral LED imaging microscope (MSLEDIM) is relatively cheaper and capable of acquiring images in reflection, transmission and scattering modes at thirteen (13) different wavelengths ranging from ultraviolet to near infrared. The microscope has been demonstrated in biomedical and entomological research fields. The MSLEDIM can be used in various scientific research fields for imaging microscopic samples.

Malaria parasite, Plasmodium falciparum, uses haemoglobin in host red blood cells (RBCs) as a major source of nutrient in ring and trophozoite stages. This brings about changes in the morphology and functional characteristics of the RBCs. We investigate malaria infected RBCs and uninfected RBCs-ring and trophozoite stages using multispectral imaging technique. Four spectral bands were found to be markers for identifying infected and uninfected RBCs: 435 nm and 660 nm were common markers for the two stages whiles 590 nm and 625 nm were markers for the ring and the trophozoite stages respectively. These four spectral bands may offer potential diagnostic markers for identifying infected and uninfected RBCs, as well as distinguishing ring and trophozoite stages.

Malaria parasites, Plasmodium falciparum (P.falciparum) infections are taking a great toll on the lives of people worldwide, especially in developing countries. Recently, haemozoin detection using optical techniques tends to provide comparable parasite densities (PDs) estimation. We conducted feasibility studies on P.falciparum infected blood (i-blood) and uninfected blood (u-blood) samples from volunteers employing laser-induced fluorescence technique for PDs estimation. Fluorescence results show high intensity in u-blood than i-blood. PeakFit analysis with Loess smoothing under Lorentzian curve shows that fluorescence peak of i-blood appears red-shifted with increasing PDs. The Lorentzian curves depict that fluorescence peak intensity ratio increases with increasing PDs in i-blood samples. This technique may be potentially applied in PDs estimation to improve malaria diagnosis.

We present results from a widely tunable (296 cm−1) Quantum Cascade laser based-sensor used in sensing water vapor and target gases found in wood smoke in the rural fishing village of Elmina, Ghana.

Cassava mosaic disease (CMD) is a major constraint to cassava production in cassava growing regions. Severity of CMD symptoms on cassava leaves is usually assessed visually using an arbitrary scale, which is semi-qualitative, and does not represent the actual surface area of diseased leaf. The objective of this study was to develop a quantitative method of assessing the severity of CMD. A combination of polarimeteric digital colour images, L*a*b* colour model and K-means clustering algorithm were used to determine the areas of CMD symptoms and healthy areas on leaves. The severity of CMD on a leaf is determined by computing the percentage of the CMD symptomatic area to the total leaf area. The analysis provides relatively fast and accurate classification of Cassava mosaic diseased leaves. The proposed method will enable plant scientists to obtain accurate and reliable data, forming the basis for better decision making.

An optical Fourier system comprising two Fourier lenses arranged in a 4 - f configuration, a CCD camera coupled to an Analogue Digital (A/D) converter and interfaced to a computer, has been found effective to image and discriminate real etched alpha tracks and artifacts on LR-115. The choice of the optical components enhanced the system's sensitivity and the field of view of detector area (1.4 cm2). The use of spatial filters in the Fourier plane deforms the tracks and introduces phase variation of the Fourier transform signal. In addition narrow bandwidth filters could not cover broader range of all the tracks. Image analysis before track counting improves signals to noise ratio and increase the selectivity of the tracks signal size. The system can handle higher track density and other types track detectors.