Transactions on Machine Intelligence2821-16937220240301Phonocardiogram Analysis for Cardiovascular Disease Screening Using K-Star Algorithm828919160110.47176/TMI.2024.82ENN.RashidianDepartment of Physiotherapy, Rehabilitation Building, Shiraz University of Medical Sciences, Shiraz, IranJournal Article20240124Cardiovascular disease stands out as one of the most prevalent health issues among the population. The accurate diagnosis and effective treatment of this disease are of paramount importance. The primary objective of this research is to propose a novel model for the automatic classification of heart sounds, specifically targeting the analysis of phonocardiograms to aid in the screening and diagnosis of cardiovascular disease. In this study, a dataset consisting of 942 samples, recorded heart sounds, each characterized by 23 features. The K-Star algorithm was employed for the classification of heart sounds. The K-Star algorithm is a model-based learning method that utilizes entropy theory as a distance measure. This approach maximizes the extraction of information from available data, offering a consistent methodology for managing both symbolic features and missing values effectively. The algorithm calculates the distance between two samples by considering the complexity of transforming one sample into another. The Waka tool was employed to implement this algorithm. Through the utilization of the K-Star algorithm, the accuracy of phonocardiogram analysis for cardiovascular disease screening was significantly enhanced, achieving a notable accuracy rate of 80.8917%. This research contributes to the development of a reliable and efficient tool for the automatic classification of heart sounds, aiding in the early detection and screening of cardiovascular diseases.https://www.tmachineintelligence.ir/article_191601_76717eca7680959d81112c802fb2531a.pdfTransactions on Machine Intelligence2821-16937220240301Tremor Suppression in Robot-Assisted Minimally Invasive Surgery using Kalman Filter Adapted by Fuzzy System and Reinforcement Learning909718739210.47176/TMI.2024.90ENM.R.Sayyed NooraniDepartment of Mechatronics Engineering, University of Tabriz, Tabriz, Iran0000-0002-7487-8233S.NaiemGraduated in Medical Engineering-Biomechanics, Department of Mechatronics Eng., University of TabrizJournal Article20240301Robot-assisted minimally invasive surgery (RA-MIS) has seen growing adoption in recent years due to its advantages in precision, reduced trauma, and shorter recovery times. A critical challenge in RA-MIS, particularly in remote leader-follower robotic configurations, is the suppression of involuntary hand tremors exhibited by surgeons. These physiological tremors, often induced by fatigue, stress, or prolonged procedures, can significantly impair surgical accuracy. To ensure optimal performance, it is essential to attenuate these unwanted vibrations during surgical tasks. One conventional solution is to model the tremor as an external noise source and apply filtering techniques such as the Kalman filter to isolate and remove the noise. However, since the characteristics of hand tremors are inherently time-varying, static filtering approaches may fall short in dynamic surgical environments. To address this, we propose two adaptive methods for enhancing the Kalman filter: one based on a fuzzy inference system, and another using a reinforcement learning technique Q-learning for real-time updating of the filter’s error covariance matrix. Simulation results indicate that both approaches significantly improve tremor suppression by dynamically adjusting to variations in the signal. These adaptive filtering techniques provide a promising solution for increasing precision and stability in robotic-assisted surgical systems.https://www.tmachineintelligence.ir/article_187392_3e0a73638fc0f26b56b25bfd64cfcfe0.pdfTransactions on Machine Intelligence2821-16937220240301Phase Transition of the 2D Square Ising Model in a Homogeneous Magnetic Field Using the Metropolis Monte Carlo Algorithm and Separation of Different Phases via CNN Method9810619160410.47176/TMI.2024.98ENR.RastgarpourFaculty of Engineering Sciences, School of Engineering, University of Tehran, Tehran, IranN.MajdAssistant Professor, Department of engineering science, college of engineering, university of Tehran, Tehran, Iran0000-0002-5652-7405Journal Article20240306Quantum spin networks represent systems in which quantum spins are arranged on a topological lattice, where the nature of spin interactions and the influence of external magnetic fields can lead to complex collective behaviors. The Hamiltonian governing such systems determines the energy landscape and phase transitions under various thermodynamic conditions. In this paper, we focus on the two-dimensional Ising spin model with periodic boundary conditions subjected to a uniform external magnetic field. Initially, we employ the Metropolis Monte Carlo (MP-MN) algorithm to simulate the system and identify its phase transitions at different magnetic field strengths. The emergence of ordered and disordered phases in response to thermal fluctuations and magnetic interactions is systematically analyzed. Subsequently, we explore the application of convolutional neural networks (CNNs), a powerful class of deep learning models, to detect and classify the phases of the Ising model based on spin configurations generated at a fixed temperature. The CNN is trained using labeled data representing different magnetic field values, and its performance in phase prediction is quantitatively evaluated. The results demonstrate that CNNs can successfully learn complex spin patterns and provide accurate classification of spin phases, highlighting their potential for analyzing phase transitions in statistical physics models.https://www.tmachineintelligence.ir/article_191604_92c29608543abd49a4b02ccc9469c2fe.pdfTransactions on Machine Intelligence2821-16937220240301Control of Vehicle Active Suspension System Using Neural Network10713519160510.47176/TMI.2024.107ENS. Z.Ahmadi SheykhShabaniMasters Student, Control Engineering, Iran University of Science and Technology, Tehran, Iran.D.SheikhiMasters Student, Control Engineering, Iran University of Science and Technology, Tehran, IranJournal Article20240307The primary cause of vehicle vibration stems from road irregularities. Addressing this, an effective strategy involves implementing a robust artificial neural network control system to manage the vehicle suspension system's vibrations. To achieve comprehensive vibration control for the entire suspension system, a robust neural network-based control system is employed. The complete vehicle system operates with 7 degrees of freedom, encompassing vertical axis motion, angular variations around the X axis, and angular changes around the Y axis of the car chassis. The proposed control system integrates a robust controller, a neural controller, and a neural network model tailored for the vehicle suspension system. To assess simulation outcomes, a proportional integral derivative (PID) controller is utilized for overall vehicle suspension system vibration control. The study introduces random road roughness as a disturbance factor applied to the proposed control system. Simulation results affirm that the suggested neural control system demonstrates highly effective control performance, with minimal error in adapting to unexpected road disturbances affecting the vehicle suspension.https://www.tmachineintelligence.ir/article_191605_4a5d2cf2fdbf7d24b72ef73a5250c03b.pdfTransactions on Machine Intelligence2821-16937220240301An overview of FPGA-based digital insulin-glucose implementations Regulator for type 2 diabetic patients13614619160610.47176/TMI.2024.136ENK. BaratiK. BaratiDepartment of Electronics, Lorestan University, Khorramabad, IranA. S.HamidiDepartment of Electronics, Lorestan University, Khorramabad, IranA.FarmaniDepartment of Electronics, Lorestan University, Khorramabad, IranJournal Article20240402This study details the development of a digital insulin-glucose regulation system for type 2 diabetes management, utilizing a Field Programmable Gate Array (FPGA) board. The system is designed to monitor and control insulin levels in patients by measuring their blood glucose levels only. Unlike other solutions that rely on general-purpose programmable hardware, this regulator is built entirely on a hardware-based architecture without the need for software, as elaborated in this document. A prototype was created to test its effectiveness in two scenarios: (i) an open loop mode where the regulator operates independently, and (ii) a closed loop mode where it functions as an artificial pancreas and is linked to a group of one hundred virtual patients. These patients were created using a detailed theoretical model approved by the U.S. Food and Drug Administration for pre-clinical trials of glucose regulation methods. The virtual patients exhibit similar patterns in glucose fluctuations with varying peak and trough levels post-meal. The outcomes of these tests are analyzed and compared with results derived from theoretical model simulations conducted in SIMULINK. The comparison shows relative errors within ±1%, indicating the high precision of this digital insulin-glucose regulation system. The hardware implementation processes each virtual patient's glucose data in approximately 1.1 μs and consumes about 36 mW of power. These promising results encourage further research into digital systems for glucose regulation that could be incorporated into very-large-scale integration (VLSI) as System-on-Chips or Lab-on-Chips. Such developments could pave the way for advanced, miniaturized devices suitable for portable, wearable, and implantable medical applications.https://www.tmachineintelligence.ir/article_191606_a3f47cc43a43648a7fb1f1f6ee6dc5e4.pdfTransactions on Machine Intelligence2821-16937220240301Designing Optimal Predictive Control Model for Boost Converter in Solar Inverters with the Help of Meta-Engineering Algorithms14716019362010.47176/TMI.2024.147END.KhalilzadeDepartment of Electrical Engineering, K. N. Toosi University of Technology, Tehran, IranA.AhrabiDepartment of Electrical Engineering, K. N. Toosi University of Technology, Tehran, IranJournal Article20240306This study focuses on modeling the DC-DC converter and implementing predictive model control methods on the target system. The research aims to compare the efficiency of this approach against classical methods and devise a strategy for maximizing power extraction. Utilizing MATLAB software, we simulate the proposed converter and control method, analyzing the obtained data and results through comparison with alternative methods. The article aims to enhance the performance of the boost converter and DC-DC converter through predictive control. Specifically, the boost converter is tasked with converting 50 V photovoltaic voltage to 110 V. Our initial focus lies on designing predictive control for the boost converter, acknowledging its potential for higher accuracy compared to other control methods. However, a notable challenge of predictive control lies in manually determining coefficients in the cost function. In this work, we address this challenge by employing amplifying coefficients on the input and output of the MPC converter and determining these values using a meta-engineering algorithm. This approach aims to refine predictive control for improved performance. The proposed control demonstrates promising accuracy and speed in reaching set point values, with favorable energy metrics. https://www.tmachineintelligence.ir/article_193620_0aa3f451cc7b54ae1c2a8f13bc651ba2.pdf