Alzheimer's Disease Detection from Brain MRI Scans Using Convolutional Neural Networks Aryan Singh Chauhan, Manikanta Kancharla, Charu Pathak, Bhargavi Devapatla 2025 International Conference on Cognitive Computing in Engineering Communications Sciences and Biomedical Health Informatics Ic3ecsbhi 2025, 2025 The rise in Alzheimer's disease cases has become a challenge to global healthcare systems, hence making it mandatory to develop proper accurate diagnostic tools. This paper introduces an aggressive deep learning model using CNNs (Convolutional neural network) to classify Alzheimer's disease into four stages: Non-Demented, Really Gentle Demented, Light Demented, and Reasonable Demented. This variant leverages the well-balanced preparation approach and novel ways of data generation to overcome class imbalance and enhance generalization altogether. The design and practice schedule with monitoring metrics like correctness, AUC, and F 1 rating ensures optimal performance and robustness. The classification ability of the proposed design was observed to be truly remarkable and surpassed conventional tactics both in the dimensions of accuracy and sensitivity. These findings expand the prospects in CNN-based systems directed at the early and accurate diagnosis of Alzheimer's, as they become relevant for medical practice and as a basis for further innovation in medical diagnostics.
Enhanced Lathe Tool Positioning Through Wavelet-Based Image Denoising Techniques Shweta Kumari, Charu Pathak, Shruti Vashist Iet Conference Proceedings, 2024 Precise tool placement is crucial in smart manufacturing and automation. In micro-machining, tiny measurements must be very accurate. Finding edges accurately is a key step for good positioning and micrometer-level measurements. Image denoising was necessary before applying edge detection algorithms to enhance edge detection accuracy. Conventional filtering methods can be employed for denoising, but the work here investigates wavelet thresholding benefits in analyzing multiple resolutions and achieving better accuracy. An in-depth assessment of the quality of the denoised images is conducted. The quantitative analysis is done using metrics such as PSNR (peak signal to noise ratio), SSIM (structural similarity index measurement), MSE (mean squar e error) and others factors in which the wavelet based denoising method's results are compared at different scales. The results from the analysis show that the db1 wavelet based on SURE denoising method can efficiently remove a lot of Gaussian noise and restore the image details and data than Coiflets and Symlets wavelet techniques.
AI-Enhanced Early Warning Systems for Natural Disaster Detection and Mitigation using Wireless Sensor Networks Piyush Charan, Mohd Maroof Siddiqui, Varun Yadav, Charu Pathak, Yogendra Narayan, Zohaib Hasan Khan 2024 2nd International Conference Computational and Characterization Techniques in Engineering and Sciences Ic3tes 2024, 2024 Each year, natural disasters result in substantial economic and human losses, underscoring the critical requirement for predictive systems capable of issuing timely alerts. Technology, particularly Artificial Intelligence (AI), plays a pivotal role in enhancing early warning systems (EWSs) for natural disasters. AI algorithms analyze extensive datasets from various sources—such as satellite imagery, weather data, and seismic activity—to deliver more precise and nearly real-time warnings for events like earthquakes, tsunamis, and floods. This paper proposes an AI-driven approach that utilizes wireless sensor networks (WSNs) strategically deployed in disaster-prone coastal areas to monitor seismic activity and oceanic disturbances. The system aims to detect anomalies indicative of tsunamis and earthquakes, offering an advanced EWS that enhances both the accuracy and timeliness of disaster warnings, thereby mitigating potential risks to vulnerable communities. Additionally, the paper explores opportunities for further refinement of these systems to better protect at-risk populations.
Dynamics Inside a Dual-Frequency Paul Trap in the Presence of Excess Micromotion Anuranjan Kansal, Varun Saxena, Charu Pathak IEEE Transactions on Plasma Science, 2023 Dynamics inside a dual-frequency Paul trap is theoretically analyzed in the presence of excess micromotion on account of stray electric field and the field due to a mismatch of potential at the electrodes. The fields contributing to the excess micromotion have a strong influence on the kinetic energy of a single-charged particle, however, their influence is not profound when the collective dynamics of many charged particles is considered inside the trap. The collective dynamics is encapsulated by a distribution function whose time average is double-humped with respect to the velocity. The spatial variation of energy is asymmetrical. The temporal fluctuations of the energy are periodic with a period of variation depending upon the ratio of the two applied RF frequencies.
Estimation of Position of Lathe Tool Using Edge Detection Technique Shweta Kumari, Charu Pathak, Shruti Vashist, Prasant Kumar Mahapatra Proceedings of 2023 International Conference on Intelligent Systems Advanced Computing and Communication Isacc 2023, 2023 Accurate positioning of tool in mechanical devices used in smart manufacturing, mechanical automation is of utmost importance. In applications like manufacturing of micro parts, an accuracy of micrometer is required. Thus, it is proposed to opt for an economic fast and accurate tool positioning system which may directly work on the images of the tool taken by a high-resolution camera. This work contributes to estimate the position of a lathe tool using image processing. The tool is considered to move in x-y plane. The movement of the tool is calculated from its origin using Euclidean distance. The coordinates of the vertices of the tool are detected using Edge detection technique. These coordinates are used to estimate the distance travelled by the tool from the origin. Prewitt and Sobel operators are applied for edge estimation and the errors obtained in both Prewitt and Sobel are compared. As a result of experimental analysis Sobel operator is found to give better results. The maximum error obtained is 272.2 µm and the minimum error obtained is 0.1 µm.
Contactless Measurement of Error in Lathe Tool Positioning Shweta Kumari, Charu Pathak, Shruti Vashist, Prasant Kumar Mahapatra 2023 3rd International Conference on Intelligent Technologies Conit 2023, 2023 This paper presents a study on the importance of correct lathe tool positioning in various applications, including mechanical, medical, and manufacturing fields. The objective of this study is to analyze tool position error and the distance travelled by the tool when the work piece is stationary, with the aim of reducing positional inaccuracy of the lathe tool. Accurate positioning is critical in applications that require precision in the micrometer range. To achieve this, various image segmentation techniques such as Graythresh, Multithresh, and Adaptthresh have been applied and evaluated for estimating the actual tool position and error in tool movement distance. Results show that Multithresh provides the least mean absolute deviation of error and the least variance of error, making it the most accurate technique for determining the tool position at the micrometer scale, compared to Graythresh and Adaptthresh.
Field Imperfection Induced Effects in a Planar Dual Frequency Paul Trap Anuranjan Kansal, Varun Saxena, Charu Pathak IEEE Transactions on Plasma Science, 2022 The article investigates the influence of electric field imperfections that manifest themselves as hexapole and octopole aberrations in a dual frequency planar Paul trap. The characteristics of the charged particle motion are analyzed by formulating equation of motion in a pseudo-potential well wherein the hexapole and octopole field effects get superimposed to yield a Duffing type of an equation. Perturbation and harmonic balance methods are applied to estimate the axial frequency shifts and the displacement of the charged particle from the axis of the trap. The axial frequency is dependent upon whereas the displacement of the charged particle from the center of the axis is independent of the variation in voltage and frequency ratio applied to the Paul trap.
