Claim Missing Document
Check
Articles

Found 37 Documents
Search

Brain tumor detection using a MobileNetV2-SSD model with modified feature pyramid network levels Hikmah, Nada Fitrieyatul; Hajjanto, Ariq Dreiki; A. Surbakti, Armand Faris; Prakosa, Nadhira Anindyafitri; Asmaria, Talitha; Sardjono, Tri Arief
International Journal of Electrical and Computer Engineering (IJECE) Vol 14, No 4: August 2024
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.11591/ijece.v14i4.pp3995-4004

Abstract

Brain tumors, a subset of these malignancies, demand accurate and efficient diagnosis. Traditional methods use non-invasive medical imaging like magnetic resonance imaging (MRI) and computed tomography (CT). Although necessary for diagnosis, manual brain MRI picture segmentation is tedious and time-consuming. Using deep learning is a promising solution. This study proposes an innovative approach for brain tumor detection, focusing on meningioma tumors. Utilizing threshold-based segmentation, the MobileNetV2 architecture, a modified feature pyramid network (FPN), and single shot MultiBox detector (SSD), our model achieves precise localization and object detection. Pre-processing techniques such as grayscale conversion, histogram equalization, and Gaussian filtering enhance the MRI image quality. Morphological operations and thresholding facilitate tumor segmentation. Data augmentation and a meticulous dataset division aid in model generalization. The architecture combines MobileNetV2 as a feature extractor, SSD for object detection, and FPN for detecting small objects. Modifications, including lowering the minimum FPN level, enhance small object detection accuracy. The proposed model achieved a recall value of around 98% and a precision value of around 89%. Additionally, the proposed model achieved approximately 93% on both the dice similarity coefficient (DSC) value and the index of similarity. Based on the promising results, our research holds significant advancements for the field of medical imaging and tumor detection.
Cost-Effective Parkinson’s Disease Diagnosis Through IoT-Based Finger Tapping and Real-Time Machine Learning Classification Arraziqi, Dwi; Sardjono, Tri Arief; Purnomo, Mauridhi Hery
Jurnal Nasional Pendidikan Teknik Informatika: JANAPATI Vol. 14 No. 1 (2025)
Publisher : Prodi Pendidikan Teknik Informatika Universitas Pendidikan Ganesha

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.23887/janapati.v14i1.86371

Abstract

Parkinson's disease (PD) is a progressive neurological condition that significantly impacts motor functions, including finger tapping (FT). This study aims to develop a cost-effective, real-time, easily implementable, IoT-enabled electronic health record (EHR)-integrated FT analysis system capable of remotely detecting PD with high accuracy. The study uses peak amplitude, the Internet of Things (IoT), and various machine learning classifiers to detect PD through FT pattern analysis on a smartphone application. K-Nearest Neighbors, Convolutional Neural Networks, Support Vector Machines, and Logistic Regression exhibited 100% accuracy, while Naïve Bayes and Decision Trees (DT) had accuracies ranging from 71% to 92%. All classifiers had an Area Under the Curve (AUC) value of 1, except DT with an AUC value of 0.75. This study introduces a novel IoT system for PD detection that demonstrates high diagnostic accuracy, cost-effectiveness, real-time monitoring capability, easy implementation, scalability for telemedicine, and accessibility to EHR during the COVID-19 pandemic. Future studies will focus on expanding the dataset.
Improving 3D Human Pose Orientation Recognition Through Weight-Voxel Features And 3D CNNs Riansyah, Moch. Iskandar; Putra, Oddy Virgantara; Rahmanti, Farah Zakiyah; Priyadi, Ardyono; Wulandari, Diah Puspito; Sardjono, Tri Arief; Yuniarno, Eko Mulyanto; Hery Purnomo, Mauridhi
EMITTER International Journal of Engineering Technology Vol 13 No 1 (2025)
Publisher : Politeknik Elektronika Negeri Surabaya (PENS)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24003/emitter.v13i1.847

Abstract

Preprocessing is a widely used process in deep learning applications, and it has been applied in both 2D and 3D computer vision applications. In this research, we propose a preprocessing technique involving weighting to enhance classification performance, incorporated with a 3D CNN architecture. Unlike regular voxel preprocessing, which uses a zero-one (binary) approach, adding weighting incorporates stronger structural information into the voxels. This method is tested with 3D data represented in the form of voxels, followed by weighting preprocessing before entering the core 3D CNN architecture. We evaluate our approach using both public datasets, such as the KITTI dataset, and self-collected 3D human orientation data with four classes. Subsequently, we tested it with five 3D CNN architectures, including VGG16, ResNet50, ResNet50v2, DenseNet121, and VoxNet. Based on experiments conducted with this data, preprocessing with the 3D VGG16 architecture, among the five architectures tested, demonstrates an improvement in accuracy and a reduction in errors in 3D human orientation classification compared to using no preprocessing or other preprocessing methods on the 3D voxel data. The results show that the accuracy and loss in 3D object classification exhibit superior performance compared to specific preprocessing methods, such as binary processing within each voxel.
Person and Activity Recognition Based on Joint Motion Features Using Deep Learning with Drone Camera Yunardi, Riky Tri; Sardjono, Tri Arief; Mardiyanto, Ronny
International Journal of Robotics and Control Systems Vol 5, No 3 (2025)
Publisher : Association for Scientific Computing Electronics and Engineering (ASCEE)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.31763/ijrcs.v5i3.1949

Abstract

The increasing demand for drone-based surveillance systems has raised significant concerns about advancements in person and activity recognition based on joint motion features within visual monitoring frameworks. This study contributes to developing deep learning models that improve surveillance systems by using RGB video data recorded by drone cameras. In this study, a framework for person and activity recognition based on 120 datasets is proposed, from drone camera-recorded videos of 10 subjects, each performing six movements: walking, running, jogging, boxing, waving, and clapping. Joint motion features, including joint positions and joint angles, were extracted and processed as one-dimensional series data. The 1D-CNN, LeNet, AlexNet, and AlexNet-LSTM architectures were developed and evaluated for classification tasks. Evaluation results show that AlexNet-LSTM outperformed the other models in person recognition, achieving a classification accuracy of 0.8544, a precision of 0.9161, a recall of 0.8575, and an F1-score of 0.8332, while AlexNet delivered superior performance in activity recognition with an accuracy of 0.8571, a precision of 0.8442, a recall of 0.8599, and an F1-score of 0.8463. The relatively small dataset size used likely favors simpler architectures like AlexNet. These findings highlight the effectiveness of joint motion features for person identification and emphasize the suitability of simpler classifier architectures for activity classification when working with small datasets.
Lung sound classification using YAMNet, neural network, and augmentation Arifin, Jaenal; Sardjono, Tri Arief; Kusuma, Hendra
IAES International Journal of Artificial Intelligence (IJ-AI) Vol 14, No 5: October 2025
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.11591/ijai.v14.i5.pp4101-4112

Abstract

Globally, lung disease occupies a significant position as one of the main contributors to mortality rates. The characteristics of human respiratory sound signals can show a wide spectrum, ranging from normal patterns to indications of lung abnormalities. The proposed lung sound classification system is based on YAMNet as a pre-trained neural network model for medical audio recognition, which is then refined using artificial neural networks (ANN). This study presents the integration of multiple datasets and advanced pre-processing approaches. A total of 1,363 lung sound recordings from Kaggle, ICBHI, and Mendeley. This reflects the variety of clinical conditions, and differences in recording devices are combined. In order to increase the diversity of lung sound signal input, the pre-processing process is carried out through several stages, including adjusting the sampling frequency to 4 kHz, segmenting for 6 seconds, signal filtering with wavelet, min–max normalization, and data augmentation using window warping, jittering, cropping, and padding. A fold cross-validation scheme is employed to comprehensively evaluate the model's effectiveness. The evaluation results indicate that the model achieves an accuracy of 93.64%, a precision of 93.60%, a recall of 93.64%, and an F1-score of 93.52%, collectively reflecting outstanding classification performance. This work may incorporate deep learning technology into clinical practice, ultimately improving diagnosis accuracy and efficiency in the hospital setting.
Optimizing YOLOv8 for Enhanced Melon Maturity Detection with Attention Mechanisms: A Case Study from Puspalebo Orchard Umar, Ubaidillah; Sardjono, Tri Arief; Kusuma, Hendra; Yani, Mohamad; Widyantara, Helmy
JOIV : International Journal on Informatics Visualization Vol 9, No 3 (2025)
Publisher : Society of Visual Informatics

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.62527/joiv.9.3.2942

Abstract

Enhancing fruit maturity detection is crucial in the agricultural industry to ensure product quality and reduce post-harvest losses. However, commonly used maturity detection methods still rely on human visual inspection, which is prone to errors and assessment variability. Challenges like lighting variations, complex backgrounds, and diverse environmental conditions often complicate accurate and efficient detection. This study aims to develop and evaluate an optimized YOLOv8 model with attention mechanisms to detect melon maturity. The dataset was obtained from Puspalebo Orchard in East Java, Indonesia, comprising over a thousand melon images divided into three subsets: 70% for training, 20% for validation, and 10% for testing. The YOLOv8 model was modified to support the integration of attention mechanisms to enhance focus on significant features and detection accuracy. Data augmentation techniques were applied to capture environmental condition variations, improving the model's robustness. Evaluation on the validation subset showed a precision of 0.979 for all classes, recall of 0.962, mAP@50 of 0.981, and mAP@50-95 of 0.941. The model also demonstrated high efficiency for real-time applications with a preprocessing time of 0.1ms, inference time of 0.9ms, and post-process time of 0.9ms per image. The results of this study show advantages in detection detail, adaptability, and real-time efficiency compared to other studies in the past five years. Some weaknesses were identified, such as implementation complexity and the need for a large dataset. The developed YOLOv8 model improves melon maturity detection performance, offering a more accurate, efficient, and adaptive solution for the agricultural industry.
DELINEATION OF ECG FEATURE EXTRACTION USING MULTIRESOLUTION ANALYSIS FRAMEWORK Nada Fitrieyatul Hikmah; Achmad Arifin; Tri Arief Sardjono
JUTI: Jurnal Ilmiah Teknologi Informasi Vol. 18, No. 2, July 2020
Publisher : Department of Informatics, Institut Teknologi Sepuluh Nopember

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.12962/j24068535.v18i2.a992

Abstract

ECG signals have very features time-varying morphology, distinguished as P wave, QRS complex, and T wave. Delineation in ECG signal processing is an important step used to identify critical points that mark the interval and amplitude locations in the features of each wave morphology. The results of ECG signal delineation can be used by clinicians to associate the pattern of delineation point results with morphological classes, besides delineation also produces temporal parameter values of ECG signals. The delineation process includes detecting the onset and offset of QRS complex, P and T waves that represented as pulse width, and also the detection of the peak from each wave feature. The previous study had applied bandpass filters to reduce amplitude of P and T waves, then the signal was passed through non-linear transformations such as derivatives or square to enhance QRS complex. However, the spectrum bandwidth of QRS complex from different patients or same patient may be different, so the previous method was less effective for the morphological variations in ECG signals. This study developed delineation from the ECG feature extraction based on multiresolution analysis with discrete wavelet transform. The mother wavelet used was a quadratic spline function with compact support. Finally, determination of R, T, and P wave peaks were shown by zero crossing of the wavelet transform signals, while the onset and offset were generated from modulus maxima and modulus minima. Results show the proposed method was able to detect QRS complex with sensitivity of 97.05% and precision of 95.92%, T wave detection with sensitivity of 99.79% and precision of 96.46%, P wave detection with sensitivity of 56.69% and precision of 57.78%. The implementation in real time analysis of time-varying ECG morphology will be addressed in the future research.