EMITTER International Journal of Engineering Technology

Lite-FBCN: Lightweight Fast Bilinear Convolutional Network for Brain Disease Classification from MRI Image

Fri, 20 Dec 2024 12:41:15 +0000

Achieving high accuracy with computational efficiency in brain disease classification from Magnetic Resonance Imaging (MRI) scans is challenging, particularly when both coarse and fine-grained distinctions are crucial. Current deep learning methods often struggle to balance accuracy with computational demands. We propose Lite-FBCN, a novel Lightweight Fast Bilinear Convolutional Network designed to address this issue. Unlike traditional dual-network bilinear models, Lite-FBCN utilizes a single-network architecture, significantly reducing computational load. Lite-FBCN leverages lightweight, pre-trained CNNs fine-tuned to extract relevant features and incorporates a channel reducer layer before bilinear pooling, minimizing feature map dimensionality and resulting in a compact bilinear vector. Extensive evaluations on cross-validation and hold-out data demonstrate that Lite-FBCN not only surpasses baseline CNNs but also outperforms existing bilinear models. Lite-FBCN with MobileNetV1 attains 98.10% accuracy in cross-validation and 69.37% on hold-out data (a 3% improvement over the baseline). UMAP visualizations further confirm its effectiveness in distinguishing closely related brain disease classes. Moreover, its optimal trade-off between performance and computational efficiency positions Lite-FBCN as a promising solution for enhancing diagnostic capabilities in resource-constrained and or real-time clinical environments.

From Waste to Power: Fly Ash-Based Silicone Anode Lithium-Ion Batteries Enhancing PV Systems

Kania Yusriani Amalia, Tresna Dewi, Rusdianasari Rusdianasari — Fri, 20 Dec 2024 13:22:43 +0000

Indonesia's high solar irradiance, averaging 4.8 kWh/m²/day, presents a significant opportunity to harness solar power to meet growing energy demands. Fly ash, abundant in Indonesia and rich in silicon dioxide (40-60% SiO₂), can be repurposed into high-value silicon anodes. The successful extraction of silicon from fly ash, increasing SiO₂ content from 49.21% to 93.52%, demonstrates the potential for converting industrial waste into valuable battery components. Combining these advanced batteries with PV systems improves overall efficiency and reliability. Energy charge and discharge experiments reveal high energy efficiency for silicon-anode batteries, peaking at 80.53% and declining to 67.67% after ten cycles. Impedance spectroscopy tests indicate that the S120 sample, with the lowest impedance values, is most suitable for high-efficiency applications. Photovoltaic (PV) system integration experiments show that while increased irradiance generally boosts power output, other factors like PV cell characteristics and load conditions also play crucial roles. In summary, leveraging Indonesia's solar potential with fly ash-based silicon anode batteries and advanced predictive analytics addresses energy and environmental challenges. This innovative approach enhances battery performance and promotes the circular economy by converting waste into high-value products, paving the way for a sustainable and efficient energy future.

Impact of Principal Component Analysis on the Performance of Machine Learning Models for the Prediction of Length of Stay of Patients

Jagriti Jagriti, Naresh Sharma, Sandeep Aggarwal — Fri, 20 Dec 2024 13:26:38 +0000

Patient inflow, limited resources, criticality of diseases and service quality factors have made it essential for the hospital administration to predict the length of stay (LOS) for inpatients as well as outpatients. An efficient and effective LOS prediction tool can improve the patient care and minimize the cost of service by increasing the efficiency of the system through optimal allocation of available resources in the hospital. For predicting patient’s LOS, machine learning (ML) models can have encouraging results. In this paper, five ML algorithms, namely linear regression, k- nearest neighbours, decision trees, random forest, and gradient boosting regression, have been used to predict the LOS for the patients admitted to the hospital with some medical history, laboratory measurements, and vital signs collected before admission. Additionally, the impact of principal component analysis (PCA) has been analyzed on the predictive performance of all ML algorithms. A five-fold cross-validation technique has been used to validate the results of proposed ML model. The results concluded that the RF and GB model performs better with score of 0.856 and 0.855 respectively among all the ML models without using PCA. However, the accuracy of all the models increased with the PCA except KNN and LR. The GB model when used with principal components has score and MSE approximate to 0.908 and 0.49 respectively compared to the model that incorporates with the original data. Additionally, PCA has an advantageous effect on the DT, RF and GB models. Therefore, LOS for new patients can be predicted effectively using the proposed tree-based RF and GB model with using PCA.

Early Detection of Ball Bearing Faults Using the Decision Tree Method

Fri, 20 Dec 2024 13:35:09 +0000

Bearings are one of the important components in the machine that functions as a holder and positions the shaft alignment radially when rotating. Statistics show that about 50% of failures in electric motors are related to bearings. Therefore, monitoring bearing performance and efficiency before damage occurs is necessary to avoid more serious damage and save repair costs. This research aims to build a classification model that can identify bearings in normal condition and 6 types of damage (inner crack, outer crack, ball crack, and a combination of both) using the HUST dataset. The model building process begins with collecting datasets, processing and extracting dataset features, building classification models and evaluating the models that have been made. A decision tree is a type of supervised machine learning used to categorize or make predictions based on how a previous set of questions were answered. The model is a form of supervised learning, meaning that the model is trained and tested on a set of data that contains the desired categorization. The results of the decision tree model that has been built are able to identify bearing damage with an accuracy of 94.47%.

Evaluating YOLOv5s and YOLOv8s for Kitchen Fire Detection: A Comparative Analysis

Norisza Dalila Ismail, Rizauddin Ramli, Mohd Nizam Ab Rahman — Mon, 20 Jan 2025 12:09:10 +0000

Accurate and timely detection of kitchen fires is crucial for enhancing safety and reducing potential damage. This paper discusses comparative analysis of two cutting-edge object detection models, YOLOv5s and YOLOv8s, focusing on each performance in the critical application of kitchen fire detection. The performance of these models is evaluated using five main key metrics including precision, F1 score, recall, mean Average Precision across various thresholds (mAP50-95) and mean Average Precision at 50 percent threshold (mAP50). Results indicate that YOLOv8s significantly outperforms YOLOv5s in several metrics. YOLOv8s achieves a recall of 0.814 and an mAP50 of 0.897, compared to YOLOv5s' recall of 0.704 and mAP50 of 0.783. Additionally, YOLOv8s attains an F1 score of 0.861 and an mAP50-95 of 0.465, whereas YOLOv5s records an F1 score of 0.826 and mAP50-95 of 0.342. However, YOLOv5s shows a higher precision of 0.952 compared to YOLOv8s' 0.914. This detailed evaluation underscores YOLOv8s as a more effective model for precise fire detection in kitchen settings, highlighting its potential for enhancing real-time fire safety systems. Additionally, by offering the future work of integration of sensors with latest YOLO involvement can further optimize efficiency and fast detection rate.

Development of DOAS System for Hazardous Methane Detection in the Near-Infrared Region

Mon, 20 Jan 2025 12:13:30 +0000

Methane (CH₄) is a powerful greenhouse gas that greatly contributes to global warming. It is also very combustible, which means it has a large danger of causing explosions. It is crucial to tackle methane emissions, especially those arising from oil and gas extraction processes like transit pipes. An area of great potential is the advancement of dependable sensors for the detection and reduction of methane leaks, with the aim of averting dangerous consequences. An open-path differential optical absorption spectroscopy (DOAS) system was described in this paper for the purpose of detecting CH₄ gas emission at a moderate temperature. An in-depth examination of the absorption lines was conducted to determine the optimal wavelength for measurement. The Near Infrared (NIR) region was identified as the most suitable wavelength for detecting methane. Multiple measurements were conducted at different integration times (1 second, 2 seconds, and 3 seconds) to ensure reliability and determine the optimal integration time for the CH₄ detection system. The DOAS system has the capability of precisely detecting methane concentrations at 1M ppm in the NIR region with a quick integration time of 2 seconds.

Implementation of Portable Ultrasound for Heart Disease Detection Using Cloud Computing-Based Machine Learning

Mon, 20 Jan 2025 12:17:44 +0000

Heart disease remains one of the leading causes of death globally, including in Indonesia. Cardiovascular disease is the leading cause of death worldwide, resulting in a significant number of fatalities. In Indonesia, access to specialized heart examination services is limited, requiring patients to visit large hospitals equipped with specialized facilities. Echocardiographic examinations using ultrasound can measure various heart parameters, such as hemodynamics, heart mass, and myocardial deformation. Portable ultrasound devices have emerged, enabling flexible and effective heart examinations. These devices capture video data of the patient's heart condition. The data undergoes image preprocessing involving median filtering, high-boost filtering, morphological operations, thresholding, and Canny filtering. Segmentation is performed using region filters, collinear filters, and triangle equations. Tracking utilizes the Optical Flow Lucas-Kanade method, and feature extraction employs Euclidean distance and trigonometric equations. The classification stage uses Support Vector Machine (SVM). Video data is transmitted via a mobile application to the cloud, where all stages from preprocessing to classification are conducted on cloud servers. The classification results are then sent back to the mobile application. The proposed model achieved an accuracy rate of 86% with a standard deviation of 0.09, indicating that the detection system performs effectively.