Embedded Files
(Volume: 3, Issue: 3)
Data Mining Approaches To Analyze Factors Causing Polypharmacy
New diseases and chronic ailments, ruthlessly affecting the people of all ages, have led to the excessive intake of medicines. This simultaneous and extra consumption of varied medicines by a single person to treat a single or multiple bodily disorder is termed as polypharmacy. Though polypharmacy can necessarily aid or improve the medication outcomes of a specific disease-related medicine, it has recently turned out to cause adverse drug reactions and drug-to-drug interactions, which severely harm the patient’s health and wealth. In fact, managing polypharmacy has been deemed as one of the key factors to ensure ‘Medication Without Harm’, as per the 3rd WHO Global Patient Safety Challenge. Generally, the adverse effects of polypharmacy majorly result from the following factors: (i) Inappropriate and non-beneficial prescription by the medical practitioner, (ii) Self-medicating nature of patients and (iii) Patients approaching different medical practitioners for different ailments because of which the drug interactions are left unnoticed. Numerous researchers have worked on detecting polypharmacy by analyzing the General Practitioners’ (GP) prescription like, the average number of prescriptions per patient, their associated costs and the number of pharmacies involved, the patients’ age or gender and many more. However, as the physicians, patients, healthcare providers, pharmacists and health policy-makers are all involved in polypharmacy, the key indicators for detecting and managing unfavorable polypharmacy are difficult to find. Three Iranian researchers, M. Moradi, M. Modarres and M. M. Sepehri, have taken initiative to apply data mining approaches for detecting polypharmacy from the GPs’ prescription data. The researchers have proposed a data mining framework that involved pre-modeling, modeling and post-modeling steps in their article in Scientia Iranica E, 29(6). Using the raw dataset from the National Center for Health Insurance Research of the Iran Health Insurance Organization, the researchers have used C4.5 and Classification And Regression Tree (CART) for spotting out the physicians’ features associated with polypharmacy. Additionally, they have used Response Surface Method and Correlation-based Feature Selection (CFS) to enhance the performance of their approach. As per the researchers, the average number of dispensed prescriptions per pharmacy was the key feature for identifying polypharmacy. Though their findings are utmost helpful in alleviating irrational polypharmacy, the researchers have used only limited data on drug counts, physician features and similar other indicators because of the difficulty in accessing them. Hence, the upcoming researchers should collect ample number of data to suggest more generalized approaches, which can improve the prescription quality and avoid life-threatening medications. Image courtesy: www.vecteezy.com
Quantum Neural Networks Improve Image Recognition & Classification Tasks
A number of complex, real-world data computations seek parallel-processing and are left unsolved by classical computers or even supercomputers. This has led to the development of Quantum computers, which operate using the laws of quantum mechanics. The quantum computers use qubits, which can simultaneously represent the states 0 and 1 of classical computers through the superposition of states and quantum entanglement, allowing faster computations of massive and complex data. In fact, the quantum computers improve machine learning by providing: (i) Parallel and speeded-up training as well as data optimization and (ii) Effective handling of high-dimensional feature space, so as to let the rapid learning of complex patterns within them. Quanvolutional Neural Networks (QuanvNN) are a kind of neural network architecture with quantum principles of operation, recently gaining interest in image classification, signal denoising and image recognition tasks in the machine learning domain. Farina Riaz, Shahab Abdulla, Hajime Suzuki, Srinjoy Ganguly, Susan Hopkins and Ravinesh C. Deo from Australia have tested whether the QuanvNN can improve the image classification accuracy using two datasets, namely, the Modified National Institute of Standards and Technology (MNIST) dataset and the Canadian Institute for Advanced Research 10 class (CIFAR-10) dataset. Finding the improvement in classification accuracy, the researchers have proposed their new model, called Neural Network with Quantum Entanglement (NNQE), in Sensors, 23(5). The researchers denied the need for parameter optimization in the quantum circuit of their model and simultaneously provided further improvement in the classification accuracy in the aforesaid datasets. However, the researchers were confused of why their proposed model had degraded performance in the more complicated German Traffic Sign Recognition Benchmark (GTSRB) dataset and claim the need for novel quantum neural network architectures with better-designed quantum circuits for classifying colored and complex data. Since quantum machine learning is a new area of research that can render better computational power, along with improved learning and generalization capabilities, the upcoming researchers can contribute to reduce the error, decoherence and scalability issues in it to support numerous applications like, image recognition and classification, recommendation systems, Natural Language Processing, secure communication and many more. Image courtesy: www.freepik.com
Deep Learning and Machine Learning For Red Phosphor Synthesis
Red phosphor materials, which emit red light on stimulation by an appropriate energy source, like UV or blue light, find vast applications in display technologies, signage and advertising, LED lighting, automotive lighting, horticulture lighting, medical and photography. The ability to provide improved power efficiency, its tunable brighter as well as vibrant emissions and longer life-expectancy in devices employing them have made the red phosphor materials to be used in extensive lighting applications. Usually, the red phosphorous materials are formed by adding manganese ions that stimulate red emissions by inducing energy transitions within the phosphor material. However, the crystal field parameters surrounding the manganese ions greatly affect the energy transitions and hence, affects the properties of red phosphor being synthesized. Though experimental investigations might allow better-synthesis of red phosphor, which is cost-effective, eco-friendly and efficiency-improved with tunable emissions of desired colors, they are time consuming with added expenditure. So, Mega Novita, Alok Singh Chauhan, Rizky Muliani Dwi Ujianti, Dian Marlina, Haryo Kusumo, Muchamad Taufiq Anwar, Michał Piasecki and Mikhail G. Brik have attempted to study that how far the Deep Learning (DL) and the Machine Learning (ML) approaches could help in synthesizing novel red phosphor materials, which can meet the demands of the emerging lighting technologies. Specifically, the researchers have dealt with energy prediction in Manganese-doped crystals in Journal of Luminescence, Elsevier, Vol. 269. Despite finding few limitations, the researchers believe that integrating DL and ML in red phosphor material research could result in rapid material discovery, enhanced material characterization, optimized performance, reduced costs and time with data-driven insights. As per the researchers, their research could certainly foster new advancements in the field of red phosphorous synthesis and suggest the future researchers to choose appropriate DL and ML approaches. Image courtesy: www.freepik.com
Machine Learning Models To Achieve Improved Traction Performance In Tractor Tyres
Tractors have a vital role in modern farming practices like, crop planting and harvesting, tilling, fertilizing, ploughing, transportation and similar other agricultural tasks. However, the efficient functioning of the tractors greatly depends on traction, meaning the adhesion between its tyres and the area on which the tyres get contact with. Since the tractors are mostly-operated on varying terrains of different soil types, traction in tractors needs special attention, as it is necessary for achieving effective acceleration, non-skidding braking performance, fuel efficiency and vehicle stability. Not only that, the traction in tyres also affect soil compaction and crop yield. Generally, the two important parameters, which govern the traction in tyres, are the contact area and the deflection characteristics of the tyres. A larger contact area and smaller tyre deflections are usually preferred for uniform load distribution and to hold the tyres on the varying contours of the terrain, reducing the tyre’s wear and rolling resistance. But this fact holds right for tube-type tyres. So, does the contact area and tyre deflections affect the traction of tubeless tyres. Of course, yes. However, determining these parameters for tubeless tyres is very challenging because of the complex and nonlinear nature of the tyre behaviour. Rajesh Yadav and Hifjur Raheman from Agricultural and Food Engineering Department, Indian Institute of Technology Kharagpur, have tried to accurately estimate the contact area and the deflection characteristics of both tubeless and tube-type bias-ply agricultural tyres. In their article in Engineering Applications of Artificial Intelligence, Elsevier, Vol. 133, the researchers have developed two Support Vector Regression (SVR) models to achieve the prediction task. The researchers have found the tubeless tyres to impart increased traction performance and reduced rolling resistance at higher loads than that of tube-type tyres. Since the researchers have stated that their models are highly influenced by normal load, inflation pressure and tyre-type in order, there are future possibilities to attain improved traction parameters at higher loads with varied machine learning models. Image courtesy: www.freepik.com
Cell’s Poroelastic Behavior Can Be Modelled Mathematically ...
Cells are the smallest structural and functional unit that form tissues, organs and eventually, the entire body. They carry the genetic information, produce energy through metabolism, coordinate various physiological processes and responses to stimuli and many more. So, understanding the cellular behavior is utmost essential for various reasons and they include: (i) To gain insight into an emerging disease based on the cellular responses; (ii) To promote drug discovery; (iii) To aid in the investigation on regenerative tissue engineering and (iv) To support various biotechnological and biomedical processes. Usually, the mechanical properties of the cell have a vital role in determining its behavior. The mechanical properties of the cell imply the cell’s nature to respond to mechanical forces or deformations and some of them are stiffness, adhesion, tension, viscoelasticity, fluidity etc. Recently, the cell has been found to exhibit poroelastic behavior, meaning that they are not just rigid structures. The cell truly exhibits a solid-like elastic property and a fluid-like porous property, when a mechanical force is applied over it. The former property is because of its cytoskeleton, which are the network of protein filaments providing structural support to the cell. In contrast, the latter property of the cell has been acquired from the cytosol fluid of its cytoplasm. There are numerous methods available in the past to evaluate the stiffness or viscoelasticity of the cells. However, there is no mathematical model that incorporates poroelastic cell deformation and Young's modulus of solid networks. S. A. Haider, G. Kumar, T. Goyal and A. Raj were the first to mathematically model the poroelastic cell deformation. In their article in Microfluidics and Nanofluidics, Springer, Vol.28, the researchers have developed two Artificial Neural Network (ANN) models to predict the Young’s modulus and the viscosity of different cell lines, like HeLa, MCF-10A and MDA MB-231. They have done the prediction by using the cell’s migration as well as the deformation characteristics, which were obtained by letting the cell lines through a constriction microchannel. Additionally, the researchers have employed a Support Vector Machine (SVM) classifier that used the initial diameter and elongation of the cell line in the constriction microchannel to classify them. Since Lab-on-a-chip devices are being highly-utilized in biotechnological and biomedical assays in recent years, incorporating novel machine learning approaches to examine the cell’s poroelastic behavior can speed up intelligent diagnostics, drug discovery and similar other biomedical applications.
Image courtesy: www.vecteezy.com
Page updated
Google Sites
Report abuse