Muhammad Fachrurrozi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Muhammad Fachrurrozi has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
- Annisa Darmawahyuni
Universitas Sriwijaya
4 shared publications - Siti Nurmaini
Universitas Sriwijaya
3 shared publications - Firdaus Firdaus
Universitas Sriwijaya
2 shared publications - Muhammad Naufal Rachmatullah
Universitas Sriwijaya
2 shared publications - Rossi Passarella
Sriwijaya University
2 shared publications - Samsuryadi Samsuryadi
Universitas Sriwijaya
2 shared publications
Latest Publications
- Open Access
- 1 February 2026
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 15 (1) , 845-859
Delineating 12-lead ECG for automated ST-elevation and ST depression detection using deep learning
ST-elevation or ST-depression are markers of an abnormal heart condition detected through an electrocardiogram (ECG) where the tracing in the ST-segment is unusually elevated above the TP-segment (baseline). Identifying the localization of the ST-segment on an ECG is difficult because even a minor change in the ST-segment can be obscured by filtering processes. The 12-lead ECG signal is a non-invasive tool in the early detection of ST-elevation based on ST- and TP-segment, with quick and accurate interpretation. This study proposes a standard 12-lead ECG delineation model using deep learning (DL). The ECG signal has been segmented to Pstart–Pend, Pend–QRSstart, QRSstart–Rpeak, Rpeak–QRSend, QRSend-Tstart, Tstart–Tend, and Tend–Pstart. The study interpreted ST-elevation or -depression using an ECG delineation approach guided by medical rules. The findings revealed that the DL model achieved an average accuracy of 99.18%, sensitivity of 92.55%,specificity of 99.55%, precision of 92.61%, and F1-score of 92.52% in limb leads. Similarly, in chest leads, the DL model attained an accuracy of 99.16%, sensitivity of 93.10%, specificity of 99.53%, precision of 93.32%, and F1-score of 93.11%. This study also validated the DL-predicted results by a cardiologist from Mohammad Hoesin Hospital, Indonesia.
- 1 February 2026
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 15 (1) , 845-859
Open AccessDelineating 12-lead ECG for automated ST-elevation and ST depression detection using deep learning
ST-elevation or ST-depression are markers of an abnormal heart condition detected through an electrocardiogram (ECG) where the tracing in the ST-segment is unusually elevated above the TP-segment (baseline). Identifying the localization of the ST-segment on an ECG is difficult because even a minor change in the ST-segment can be obscured by filtering processes. The 12-lead ECG signal is a non-invasive tool in the early detection of ST-elevation based on ST- and TP-segment, with quick and accurate interpretation. This study proposes a standard 12-lead ECG delineation model using deep learning (DL). The ECG signal has been segmented to Pstart–Pend, Pend–QRSstart, QRSstart–Rpeak, Rpeak–QRSend, QRSend-Tstart, Tstart–Tend, and Tend–Pstart. The study interpreted ST-elevation or -depression using an ECG delineation approach guided by medical rules. The findings revealed that the DL model achieved an average accuracy of 99.18%, sensitivity of 92.55%,specificity of 99.55%, precision of 92.61%, and F1-score of 92.52% in limb leads. Similarly, in chest leads, the DL model attained an accuracy of 99.16%, sensitivity of 93.10%, specificity of 99.53%, precision of 93.32%, and F1-score of 93.11%. This study also validated the DL-predicted results by a cardiologist from Mohammad Hoesin Hospital, Indonesia.
- Open Access
- 3 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 205915-205931
Robust SegFormer-Based Cervical Neoplastic Lesion Segmentation Using Lesion-Specific Augmentation and Anatomical Label Optimization
Accurate segmentation of cervical lesions in Visual Inspection with Acetic Acid (VIA) images is essential for early cervical cancer detection, yet remains difficult due to low resolution, uneven illumination, translucent textures, and localized flare. This study proposes a multi-label segmentation framework using the SegFormer architecture, enhanced through anatomical label refinement and Lesion-Specific Augmentation (LSA). We evaluated three label configurations (5-class, 4-class, 3-class) using CNN and ViT-based models. The optimal configuration consisted of three anatomical regions: Columnar Area (Col), Lesion (L), and Ostium Uteri Externa (OUE), the latter being a clinically important structure rarely addressed in previous studies. A novel augmentation method, Opacity Modulation (OM), was introduced to simulate real-world contrast variation, complemented by Random Contour Warping (RCW) and Localized Flare/Noise Injection (LFNI). The SegFormer with OM achieved a lesion IoU of 0.6122, Dice of 0.7595, overall sensitivity of 92.64%, and balanced accuracy of 95.09%, outperforming prior VIA-based segmentation methods that primarily used binary or two-class protocols. The model was further evaluated on an independent external cohort, showing consistent performance trends and promising generalization across datasets. These findings indicate that combining anatomical label optimization with lesion-specific augmentation improves segmentation consistency and anatomical precision in VIA-based cervical imaging. The proposed framework provides a solid foundation for future studies aimed at clinically validated and reproducible AI-assisted screening research.
- 3 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 205915-205931
Open AccessRobust SegFormer-Based Cervical Neoplastic Lesion Segmentation Using Lesion-Specific Augmentation and Anatomical Label Optimization
Accurate segmentation of cervical lesions in Visual Inspection with Acetic Acid (VIA) images is essential for early cervical cancer detection, yet remains difficult due to low resolution, uneven illumination, translucent textures, and localized flare. This study proposes a multi-label segmentation framework using the SegFormer architecture, enhanced through anatomical label refinement and Lesion-Specific Augmentation (LSA). We evaluated three label configurations (5-class, 4-class, 3-class) using CNN and ViT-based models. The optimal configuration consisted of three anatomical regions: Columnar Area (Col), Lesion (L), and Ostium Uteri Externa (OUE), the latter being a clinically important structure rarely addressed in previous studies. A novel augmentation method, Opacity Modulation (OM), was introduced to simulate real-world contrast variation, complemented by Random Contour Warping (RCW) and Localized Flare/Noise Injection (LFNI). The SegFormer with OM achieved a lesion IoU of 0.6122, Dice of 0.7595, overall sensitivity of 92.64%, and balanced accuracy of 95.09%, outperforming prior VIA-based segmentation methods that primarily used binary or two-class protocols. The model was further evaluated on an independent external cohort, showing consistent performance trends and promising generalization across datasets. These findings indicate that combining anatomical label optimization with lesion-specific augmentation improves segmentation consistency and anatomical precision in VIA-based cervical imaging. The proposed framework provides a solid foundation for future studies aimed at clinically validated and reproducible AI-assisted screening research.
- Open Access
- 1 October 2025
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 14 (5) , 3728-3740
Text clustering for analyzing scientific article using pre-trained language model and k-means algorithm
Text clustering is a technique in data mining that can be used for analyzing scientific articles. In Indonesia-accredited journals, SINTA, there are two languages used, Indonesian and English. This is the first research focusing on clustering Indonesian and English texts into one cluster. In this research, bidirectional encoder representations from transformers (BERT) and IndoBERT are used to represent text data into fixed feature vectors. BERT and IndoBERT are pre-trained language models (PLMs) that can produce vector representations that take care of the position and context in a sentence. To cluster the articles, the K-Means algorithm is implemented. This algorithm has good convergence and adapts to the new examples, which helps in improved clustering performance. The best k-value in the K-Means algorithm is defined by using the silhouette score, the elbow method, and the Davies-Bouldin index (DBI). The experiment shows that the silhouette score can produce the most optimal k-value in clustering the articles, which has a mean score of 0.597. The mean score for the elbow method is 0.425, and for the DBI is 0.412. Therefore, the silhouette score optimizes the performance of PLMs and the K-Means algorithm in analyzing scientific articles to determine whether in scope or out of scope.
- 1 October 2025
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 14 (5) , 3728-3740
Open AccessText clustering for analyzing scientific article using pre-trained language model and k-means algorithm
Text clustering is a technique in data mining that can be used for analyzing scientific articles. In Indonesia-accredited journals, SINTA, there are two languages used, Indonesian and English. This is the first research focusing on clustering Indonesian and English texts into one cluster. In this research, bidirectional encoder representations from transformers (BERT) and IndoBERT are used to represent text data into fixed feature vectors. BERT and IndoBERT are pre-trained language models (PLMs) that can produce vector representations that take care of the position and context in a sentence. To cluster the articles, the K-Means algorithm is implemented. This algorithm has good convergence and adapts to the new examples, which helps in improved clustering performance. The best k-value in the K-Means algorithm is defined by using the silhouette score, the elbow method, and the Davies-Bouldin index (DBI). The experiment shows that the silhouette score can produce the most optimal k-value in clustering the articles, which has a mean score of 0.597. The mean score for the elbow method is 0.425, and for the DBI is 0.412. Therefore, the silhouette score optimizes the performance of PLMs and the K-Means algorithm in analyzing scientific articles to determine whether in scope or out of scope.
- Open Access
- 5 March 2025
- Published by Universitas Bumigora in MATRIK : Jurnal Manajemen, Teknik Informatika dan Rekayasa Komputer
- Vol. 24 (2) , 371-380
A Image Classification of Poisonous Plants Using the MobileNetV2 Convolutional Neural Network Model Method
Poisonous plants can be dangerous for many people, but some can be used as medicines or as pest killers. Some people, especially those in environments with a wide variety of plants, can take advantage of this poisonous plant. Lack of knowledge and information causes the use of this poisonous plant to be inappropriate. This research aims to develop software to classify images of poisonous plants using the Convolutional Neural Network method with the MobileNetV2 model and to compare the accuracy of classification results with various dataset configurations and varying parameters. The research method used is a Convolutional Neural Network, which has relatively high accuracy in classifying various digital images. The data used in this research consists of eight poisonous plants and several non-poisonous plants. The research results on 153 test data show that the accuracy value was 99.34%, precision was 99%, recall was 99%, and F1-Score was 99%. This research contributes to developing software that can quickly provide information and knowledge about poisonous plants, offering a high-accuracy solution for classifying poisonous plants using image data. Furthermore, implementing MobileNetV2 provides an efficient and lightweight model suitable for deployment on mobile devices, enhancing accessibility and usability in the field. The potential applications of this software extend beyond individual use, potentially benefiting agricultural, medical, and educational sectors. Future work will expand the dataset to include more plant species and refine the model to improve its robustness against diverse environmental conditions and image qualities.
- 5 March 2025
- Published by Universitas Bumigora in MATRIK : Jurnal Manajemen, Teknik Informatika dan Rekayasa Komputer
- Vol. 24 (2) , 371-380
Open AccessA Image Classification of Poisonous Plants Using the MobileNetV2 Convolutional Neural Network Model Method
Poisonous plants can be dangerous for many people, but some can be used as medicines or as pest killers. Some people, especially those in environments with a wide variety of plants, can take advantage of this poisonous plant. Lack of knowledge and information causes the use of this poisonous plant to be inappropriate. This research aims to develop software to classify images of poisonous plants using the Convolutional Neural Network method with the MobileNetV2 model and to compare the accuracy of classification results with various dataset configurations and varying parameters. The research method used is a Convolutional Neural Network, which has relatively high accuracy in classifying various digital images. The data used in this research consists of eight poisonous plants and several non-poisonous plants. The research results on 153 test data show that the accuracy value was 99.34%, precision was 99%, recall was 99%, and F1-Score was 99%. This research contributes to developing software that can quickly provide information and knowledge about poisonous plants, offering a high-accuracy solution for classifying poisonous plants using image data. Furthermore, implementing MobileNetV2 provides an efficient and lightweight model suitable for deployment on mobile devices, enhancing accessibility and usability in the field. The potential applications of this software extend beyond individual use, potentially benefiting agricultural, medical, and educational sectors. Future work will expand the dataset to include more plant species and refine the model to improve its robustness against diverse environmental conditions and image qualities.
- Open Access
- 1 June 2024
- Published by Institute of Advanced Engineering and Science in IAES International Journal of Artificial Intelligence (IJ-AI)
- Vol. 13 (2) , 1679-1686
Improving the performance for automated brain tumor classification on magnetic resonance imaging deep learningbased
Brain tumor is an uncontrolled growth of abnormal cell in the brain. Early diagnosis of brain tumor has a crucial step in this type of cancer, which is fatal. Magnetic resonance imaging (MRI) is one of the examination tools to examine brain anatomy in clinical practice. The high resolution and clear separation of the tissue enable medical experts to identify brain tumor. The earlier of brain tumor is detected, the wider of treatment options. However, manually analysed of brain anatomy on MRI images are time-consuming. Computer-aided diagnosis with automated way is helpful solution to help management with unreliable degrees of automation to trace various tissue boundaries. This study proposes convolutional neural network (CNN) with its excellences to automated features extraction in convolution layer. The popular architectures of CNN, i.e., visual geometry group16 (VGG16), residual network-50 (resNet-50), inceptionV3, mobileNet, and efficientNetB7 in medical image processing are compared to brain tumor classification task. As the results, VGG16 outperformed other architectures of CNN in this study. VGG16 yields 100% accuracy, precision, sensitivity, specificity, and F1-score for testing set data. The results show the excellent performance in classifying brain tumor and no tumor from MRI images that demonstrate the efficiency of system suggested.
- 1 June 2024
- Published by Institute of Advanced Engineering and Science in IAES International Journal of Artificial Intelligence (IJ-AI)
- Vol. 13 (2) , 1679-1686
Open AccessImproving the performance for automated brain tumor classification on magnetic resonance imaging deep learningbased
Brain tumor is an uncontrolled growth of abnormal cell in the brain. Early diagnosis of brain tumor has a crucial step in this type of cancer, which is fatal. Magnetic resonance imaging (MRI) is one of the examination tools to examine brain anatomy in clinical practice. The high resolution and clear separation of the tissue enable medical experts to identify brain tumor. The earlier of brain tumor is detected, the wider of treatment options. However, manually analysed of brain anatomy on MRI images are time-consuming. Computer-aided diagnosis with automated way is helpful solution to help management with unreliable degrees of automation to trace various tissue boundaries. This study proposes convolutional neural network (CNN) with its excellences to automated features extraction in convolution layer. The popular architectures of CNN, i.e., visual geometry group16 (VGG16), residual network-50 (resNet-50), inceptionV3, mobileNet, and efficientNetB7 in medical image processing are compared to brain tumor classification task. As the results, VGG16 outperformed other architectures of CNN in this study. VGG16 yields 100% accuracy, precision, sensitivity, specificity, and F1-score for testing set data. The results show the excellent performance in classifying brain tumor and no tumor from MRI images that demonstrate the efficiency of system suggested.
- Open Access
- 14 June 2023
- Published by Acadlore Publishing Services Limited in Acadlore Transactions on Applied Mathematics and Statistics
- Vol. 1 (1) , 33-43
Evaluating the Influence of the International Civil Aviation Organization on Aircraft Accident Rates and Fatalities: A Seven-Decade Historical Data Analysis
The advent of air travel, originally proposed by the Wright brothers, has led to a significant surge in aircraft usage for human transportation. In its nascent stages, this mode of transport was linked with a high frequency of accidents and consequent fatalities, placing it in the high-risk category. To counter these risks, the International Civil Aviation Organization (ICAO) was established in 1947 as a collaborative effort among numerous countries with the primary goal of enhancing aviation safety regulations. This study analyzed archival data from the Bureau of Aircraft Accidents Archives (B3A), covering a span of 72 years from 1918, the year of the first commercial airplane crash, until 2020. The objective was to understand the ICAO's impact on altering accident rates, fatalities, and underlying causes. Analytical methodologies encompassed both descriptive statistics—examining data distribution, central tendencies, and category frequencies—and exploratory data analysis (EDA) to identify variable relationships and outlier identification. The results indicated that ICAO's interventions have led to a notable decline in accident rates, with an annual average reduction of 70.9%, and a corresponding decrease in incidents attributed to technical factors. However, an unexpected trend was the increase in fatalities despite the drop in accident numbers, attributable to the introduction of larger aircraft designs carrying more passengers per flight. The findings underscore the ICAO's successful efforts in reducing aircraft accidents, but also suggest a need for further exploration into factors contributing to the rise in fatalities.
- 14 June 2023
- Published by Acadlore Publishing Services Limited in Acadlore Transactions on Applied Mathematics and Statistics
- Vol. 1 (1) , 33-43
Open AccessEvaluating the Influence of the International Civil Aviation Organization on Aircraft Accident Rates and Fatalities: A Seven-Decade Historical Data Analysis
The advent of air travel, originally proposed by the Wright brothers, has led to a significant surge in aircraft usage for human transportation. In its nascent stages, this mode of transport was linked with a high frequency of accidents and consequent fatalities, placing it in the high-risk category. To counter these risks, the International Civil Aviation Organization (ICAO) was established in 1947 as a collaborative effort among numerous countries with the primary goal of enhancing aviation safety regulations. This study analyzed archival data from the Bureau of Aircraft Accidents Archives (B3A), covering a span of 72 years from 1918, the year of the first commercial airplane crash, until 2020. The objective was to understand the ICAO's impact on altering accident rates, fatalities, and underlying causes. Analytical methodologies encompassed both descriptive statistics—examining data distribution, central tendencies, and category frequencies—and exploratory data analysis (EDA) to identify variable relationships and outlier identification. The results indicated that ICAO's interventions have led to a notable decline in accident rates, with an annual average reduction of 70.9%, and a corresponding decrease in incidents attributed to technical factors. However, an unexpected trend was the increase in fatalities despite the drop in accident numbers, attributable to the introduction of larger aircraft designs carrying more passengers per flight. The findings underscore the ICAO's successful efforts in reducing aircraft accidents, but also suggest a need for further exploration into factors contributing to the rise in fatalities.
- Open Access
- 1 June 2015
- Published by Universitas Ahmad Dahlan in TELKOMNIKA (Telecommunication Computing Electronics and Control)
- Vol. 13 (2) , 597-603
Iris Image Recognition Based on Independent Component Analysis and Support Vector Machine
Iris has a very unique texture and pattern, different for each individual and the pattern will remain stable, making it possible as biometric technology called iris recognition. In this paper, 150 iris image from Dept. Computer Science, Palacky University in Olomouc iris database used for iris recognition based on independent component analysis and support vector machine. There are three steps for developing this research namely, image preprocessing, feature extraction and recognition. First step is image preprocessing in order to get the iris region from eye image. Second is feature extraction by using independent component analysis in order to get the feature from iris image. Support vector machine (SVM) is used for iris classification and recognition. In the end of this experimental, the implement method will evaluated based upon Genuine Acceptance Rate (GAR). Experimental result shown that the recognize rate from variation of training data is 52% with one data train, 73% with two data train and 90% three data train. From experimental result also shows that this technique produces good performance.
- 1 June 2015
- Published by Universitas Ahmad Dahlan in TELKOMNIKA (Telecommunication Computing Electronics and Control)
- Vol. 13 (2) , 597-603
Open AccessIris Image Recognition Based on Independent Component Analysis and Support Vector Machine
Iris has a very unique texture and pattern, different for each individual and the pattern will remain stable, making it possible as biometric technology called iris recognition. In this paper, 150 iris image from Dept. Computer Science, Palacky University in Olomouc iris database used for iris recognition based on independent component analysis and support vector machine. There are three steps for developing this research namely, image preprocessing, feature extraction and recognition. First step is image preprocessing in order to get the iris region from eye image. Second is feature extraction by using independent component analysis in order to get the feature from iris image. Support vector machine (SVM) is used for iris classification and recognition. In the end of this experimental, the implement method will evaluated based upon Genuine Acceptance Rate (GAR). Experimental result shown that the recognize rate from variation of training data is 52% with one data train, 73% with two data train and 90% three data train. From experimental result also shows that this technique produces good performance.