Papers by F M Javed Mehedi Shamrat
IETI Transactions on Data Analysis and Forecasting (iTDAF), 2025
Humans moving in crowded spaces adapt their pace or alter their initial path. A variety of socio... more Humans moving in crowded spaces adapt their pace or alter their initial path. A variety of sociocultural factors and personal preferences influence these interactions. With the development of autonomous moving platforms that need to share their physical environment with humans, this task has become increasingly valuable. In this paper, we present a way to conveniently evaluate the ability of a model to predict social interactions between pedestrians. In this manner, we conduct experiments on two prediction models: Social long short-term memory (LSTM) and Vanilla LSTM models. We use real datasets by generating synthetic datasets for which we can define the impact of social interactions on the motion of pedestrians. These hand-tailored datasets exclude individual interactions and focus on the interactions between individuals. By comparing the models’ performances on these datasets, we show that while the Social LSTM model can predict social interactions, the Vanilla LSTM model cannot. For our analysis, we introduce evaluation metrics that focus on the interactions between pedestrians, and these metrics go beyond the commonly used average and final displacement error for trajectory prediction. In particular, we analyze the prediction errors in regions of trajectories highly influenced by social interactions. Furthermore, we analyze the collision behavior of the model’s predictions and classify trajectories concerning their degree of non-linearity. This allows us to compare the models’ performances on motions differently influenced by social interactions.
PeerJ Computer Science, 2024
Gastrointestinal (GI) diseases are prevalent medical conditions that require accurate and timely ... more Gastrointestinal (GI) diseases are prevalent medical conditions that require accurate and timely diagnosis for effective treatment. To address this, we developed the Multi-Fusion Convolutional Neural Network (MF-CNN), a deep learning framework that strategically integrates and adapts elements from six deep learning models, enhancing feature extraction and classification of GI diseases from endoscopic images. The MF-CNN architecture leverages truncated and partially frozen layers from existing models, augmented with novel components such as Auxiliary Fusing Layers (AuxFL), Fusion Residual Block (FuRB), and Alpha Dropouts (αDO) to improve precision and robustness. This design facilitates the precise identification of conditions such as ulcerative colitis, polyps, esophagitis, and healthy colons. Our methodology involved preprocessing endoscopic images sourced from open databases, including KVASIR and ETIS-Larib Polyp DB, using adaptive histogram equalization (AHE) to enhance their quality. The MF-CNN framework supports detailed feature mapping for improved interpretability of the model's internal workings. An ablation study was conducted to validate the contribution of each component, demonstrating that the integration of AuxFL, αDO, and FuRB played a crucial part in reducing overfitting and efficiency saturation and enhancing overall model performance. The MF-CNN demonstrated outstanding performance in terms of efficacy, achieving an accuracy rate of 99.25%. It also excelled in other key performance metrics with a precision of 99.27%, a recall of 99.25%, and an F1-score of 99.25%. These metrics confirmed the model's proficiency in accurate classification and its capability to minimize false positives and negatives across all tested GI disease categories. Furthermore, the AUC values were exceptional, averaging 1.00 for both test and validation sets, indicating perfect discriminative ability. The findings of the P-R curve analysis and confusion matrix further confirmed the robust classification performance of the MF-CNN. This research introduces a technique for medical imaging that can potentially transform diagnostics in gastrointestinal healthcare facilities worldwide.
IEEE, 2024
Otitis media represents inflammation or infection of the middle ear, often stemming from colds, s... more Otitis media represents inflammation or infection of the middle ear, often stemming from colds, sore throats, or respiratory infections. To address diagnostic challenges, this study introduces the EARnet-AR method, a novel advanced approach for automatically classifying otoscope tympanic membrane images. We employed 1100 otoscopic augmented images were utilized to conduct this research. To enhance the clarity and accuracy of otoscopic images, our research incorporated median filtering, an effective noise reduction technique, and gamma correction to adjust the image's luminance, ensuring a more consistent visual representation. In our initial evaluations, we tested the capabilities of five renowned pre-trained models using a comprehensive set of evaluation metrics, including accuracy, sensitivity, specificity, precision, f1-score, FPR, and FNR. The confusion matrices also detailed the model's capabilities in classifying each specific ear condition. Moreover, our proposed novel model EARnet-AR, enriched with Attention and RNN features, surpasses the other models, achieving training, testing and validation accuracies of 97.5%, 96.0%, and 96.5%, respectively. With such precise diagnostic tools, medical professionals can potentially achieve more timely and accurate treatments, enhancing patient care and outcomes in otitis media cases.
Healthcare Analytics, 2024
Diabetic retinopathy (DR) involves retina damage due to diabetes, often leading to blindness. It ... more Diabetic retinopathy (DR) involves retina damage due to diabetes, often leading to blindness. It is diagnosed via color fundus injections, but the manual analysis is cumbersome and error-prone. While computer vision techniques can predict DR stages, they are computationally intensive and struggle with complex data extraction. In this research, our prime objective was to automate the process of DR classification into its various stages using convolutional neural network (CNN) models. We employed the performance of fifteen pre-trained models with our novel proposed diabetic retinopathy network (DRNet13) model. We aimed to discern the most efficient model for accurate diabetic retinopathy (DR) staging based on fundus images from five DR classes. We preprocessed the image using a median filter for noise reduction and Gamma correction for image enhancement. We expanded our dataset from 3662 to 7500 images to create a more generalized training model through various augmentation techniques. We also evaluated multiple evaluation metrics, including accuracy, precision, F1score, Sensitivity, Specificity, Area under the curve (AUC), Mean Squared Error (MSE), False Positive Rate (FPR), False Negative Rate (FNR), in addition to confusion matrices for an in-depth comparison of the performance of these models. Feature maps were employed to illuminate decision making areas in the DRNet13 model, which achieved a 97 % accuracy rate for DR detection, surpassing other CNN architectures in speed and efficiency. Despite a few misclassifications, the model's capability to identify critical features demonstrates its potential as an impactful diagnostic tool for timely and accurate identification of diabetic retinopathy.
Retinal disorders such as age-related macular degeneration and diabetic macular edema can lead to... more Retinal disorders such as age-related macular degeneration and diabetic macular edema can lead to permanent blindness. Optical coherence tomography (OCT) enables professionals to observe cross-sections of the retina, which aids in diagnosis. Manually analyzing images is time-consuming, difficult, and prone to mistakes. In the dynamic and constantly evolving domain of artificial intelligence (AI) and medical imaging, our research represents a significant development in the field of retinal diagnostics. In this study, we introduced "RetiNet", an advanced hybrid model that is derived from the best features of ResNet50 and DenseNet121. To the model, we utilized an open-source retinal dataset that underwent a meticulous refinement process using a series of preprocessing techniques. The techniques involved Histogram Equalization for the purpose of achieving optimal contrast, Gaussian blur to mitigate noise, morphological operations to facilitate precise feature extraction, and Data Balancing to ensure impartial model training. These operations led to the attainment of a test accuracy of 98.50% by RetiNet, surpassing the performance standard set by existing models. A web application has been developed with the purpose of disease prediction, providing doctors with assistance in their diagnostic procedures. Through the development of RetiNet, our research not only transforms the accuracy of retinal diagnostics but also introduces an innovative combination of deep learning and application-oriented solutions. This innovation brings in a novel era characterized by improving reliability and efficiency in the field of medical diagnostics.
Scientific Reports, 2023
Heart failure (HF) is a leading cause of mortality worldwide. Machine learning (ML) approaches ha... more Heart failure (HF) is a leading cause of mortality worldwide. Machine learning (ML) approaches have shown potential as an early detection tool for improving patient outcomes. Enhancing the effectiveness and clinical applicability of the ML model necessitates training an efficient classifier with a diverse set of high-quality datasets. Hence, we proposed two novel hybrid ML methods ((a) consisting of Boosting, SMOTE, and Tomek links (BOO-ST); (b) combining the best-performing conventional classifier with ensemble classifiers (CBCEC)) to serve as an efficient early warning system for HF mortality. The BOOST was introduced to tackle the challenge of class imbalance, while CBCEC was responsible for training the processed and selected features derived from the Feature Importance (FI) and Information Gain (IG) feature selection techniques. We also conducted an explicit and intuitive comprehension to explore the impact of potential characteristics correlating with the fatality cases of HF. The experimental results demonstrated the proposed classifier CBCEC showcases a significant accuracy of 93.67% in terms of providing the early forecasting of HF mortality. Therefore, we can reveal that our proposed aspects (BOO-ST and CBCEC) can be able to play a crucial role in preventing the death rate of HF and reducing stress in the healthcare sector.
International Journal of Advanced Computer Science and Applications, Nov 30, 2023
Brain tumors originating from uncontrolled growth of abnormal cells in the brain, presents a sign... more Brain tumors originating from uncontrolled growth of abnormal cells in the brain, presents a significant challenge in healthcare due to their various symptoms and infrequency. While Magnetic Resonance Imaging (MRI) is essential for accurately identifying and diagnosing malignant tumors, manual interpretation is often complex and sensitive to mistakes. To address this, we introduce BrainTumorNet, a specialized convolutional neural network (CNN) created for MRI-based brain tumor diagnosis. We ensure improved image quality and a robust dataset for model training by including preprocessing approaches involving CLAHE and data augmentation. Additionally, we integrated a blockchain-based data retrieval technology to enhance the security, traceability, and collaboration in MRI data management across several medical institutions. This blockchain framework ensures that MRI data, once input from hospitals, stays immutable and can be safely retrieved based on unique hospital IDs, promoting a trustable environment for data exchange. Performance assessments conducted on multiple MRI datasets showcased BrainTumorNet's commendable proficiency, with accuracy rates of 98.66%, 97.17% and 94.24% on the dataset 1, dataset 2, and dataset 3, respectively. The model's performance was evaluated using a comprehensive set of metrics, including accuracy, specificity, recall, precision, f1-score, and confusion matrix. These measures are essential for evaluating a model's strengths and limits, emphasizing BrainTumorNet's ability to generate accurate and relevant predictions and its effectiveness in determining negative classification. BrainTumorNet's performance was compared with six renowned deep learning architectures: VGG16, ResNet50, AlexNet, MobileNetV2, InceptionV3, and DenseNet121. Our work highlights BrainTumorNet's potential capabilities in simplifying and boosting the accuracy of MRI-based brain tumor diagnosis while ensuring data integrity and collaboration through blockchain.
An Intelligent Thyroid Diagnosis System Utilising Multiple Ensemble and Explainable Algorithms with Medical Supported Attributes
IEEE Transactions on Artificial Intelligence, 2023
The widespread impact of thyroid disease and its diagnosis is a challenging task for healthcare e... more The widespread impact of thyroid disease and its diagnosis is a challenging task for healthcare experts. The conventional technique for predicting such a vital disease is complex and time-consuming. A data-driven approach may offer predictive solutions, but it relies on all relevant attributes, which are computationally expensive. Hence, we propose a novel machine learning (ML) based disease prediction system that could potentially predict it by considering three crucial steps. First, to reduce the dimension of the dataset, three feature selection techniques were employed, including Feature Importance (FIS), Information Gain Selections (IGS), and Least Absolute Shrinkage and Selection Operator (LAS). Moreover, recommended medical references were considered while developing a feature set having the identical attributes as High-Risk Factors (HRF). Second, the models, including the Three Stage Hybrid Classifier ( 3SHC ) and the Three Stage Hybrid Artificial Neural Network ( 3SHANN ), are used as classifiers on the training data set. Third, a Local Interpretable Model-agnostic Explanations (LIME) to the 3SHC with the HRF samples was applied to individually explain the predictions. Then, the overall behaviors of both gender and age categories were explored with the help of a Partial Dependence Plot (PDP). Finally, the proposed system is validated with extensive experiments where the 3SHC achieves an accuracy (ACC) of 99.29%, which can play a crucial role in preventing thyroid disease and alleviating stress in the healthcare sector.
Bulletin of Electrical Engineering and Informatics, 2023
Cauliflower disease is a primary cause of reduced cauliflower yield. Preventing cauliflower disea... more Cauliflower disease is a primary cause of reduced cauliflower yield. Preventing cauliflower disease requires early diagnosis. In the scope of this study, we suggested an agro-medical expert system that would make it easier to diagnose cauliflower disease. In this method, a digital image must be taken off the phone or handled device to diagnose cauliflower sickness. A data augmentation technique was initially used to construct a vast data set. The disease-affected parts of the cauliflower were then segmented using kmeans clustering. Following that, ten statistical and gray-level co-occurrence matrix (GLCM) features were retrieved from the segmented pictures. After choosing the top n features (N ranged from 5 to 10), the synthetic minority oversampling technique (SMOTE) approach was used to handle training datasets with different amounts of each feature. After that, we utilized five machine learning (ML) algorithms and evaluated their performance using seven performance evaluation matrices for both augmented and nonaugmented datasets. The same procedure was performed on both datasets. Then, we use both datasets to test how well the classifier works. Logistic regression (LR) is the most accurate method for the top nine features in the augmented dataset (90.77%).
IEEE Access, 2023
Searching, retrieving, and arranging text in ever-larger document collections necessitate more ef... more Searching, retrieving, and arranging text in ever-larger document collections necessitate more efficient information processing algorithms. Document categorization is a crucial component of various information processing systems for supervised learning. As the quantity of documents grows, the performance of classic supervised classifiers has deteriorated because of the number of document categories. Assigning documents to a predetermined set of classes is called text classification. It is utilized extensively in a wide range of data-intensive applications. However, the fact that real-world implementations of these models are plagued with shortcomings begs for more investigation. Imbalanced datasets hinder the most prevalent high-performance algorithms. In this paper, we propose an approach name multi-class Convolutional Neural Network (MCNN)-Long Short-Time Memory (LSTM), which combines two deep learning techniques, Convolutional Neural Network (CNN) and Long Short-Time Memory, for text classification in news data. CNN's are used as feature extractors for the LSTMs on text input data and have the spatial structure of words in a sentence, paragraph, or document. The dataset is also imbalanced, and we use the Tomek-Link algorithm to balance the dataset and then apply our model, which shows better performance in terms of F1score (98%) and Accuracy (99.71%) than the existing works. The combination of deep learning techniques used in our approach is ideal for the classification of imbalanced datasets with underrepresented categories. Hence, our method outperformed other machine learning algorithms in text classification by a large margin. We also compare our results with traditional machine learning algorithms in terms of imbalanced and balanced datasets.
A Convolutional Neural Network Based Classification Approach for Breast Cancer Detection
IEEE, 2023
Thousands of women worldwide are diagnosed with breast cancer yearly, which may be fatal if not t... more Thousands of women worldwide are diagnosed with breast cancer yearly, which may be fatal if not treated. The diagnosis of the condition may take years, by which time the patient has little choice except to have the affected breast removed. Early diagnosis and treatments are the best ways to stop this disease's spread. In this study, the authors presented a Computer Aided Diagnosis (CAD) system to assist in breast cancer diagnosis. The study uses the Wisconsin breast cancer dataset to classify benign and malignant data. For the classification, three pre-trained Deep learning algorithms: Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Multilayer Perceptron (MLP), were used. A novel CNN model that exceeds the performance efficiency of three pre-trained models and requires minimal compilation time is proposed. A number of evaluation matrices are used to analyze the models' classification abilities. Upon closer inspection, it has been established that the proposed CNN model outperforms CNN, LSTM, and MLP models with validation accuracy of 97.85%. CNN and LSTM performed with accuracies of 94.12% with the Adagrad optimizer and 93.5% with the Adam optimizer, respectively. Furthermore, MLP performance with 92.44% accuracy using the Adam optimizer. The proposed CNN model achieves the lowest Loss value and compilation time. In addition, the models' recall value, precision, and f1-score are computed to pick out the most effective model for diagnosing breast cancer on numeric data.
Bulletin of Electrical Engineering and Informatics, 2023
According to the American cancer society, breast cancer is one of the leading causes of women's m... more According to the American cancer society, breast cancer is one of the leading causes of women's mortality worldwide. Early identification and treatment are the most effective approaches to halt the spread of this cancer. The objective of this article is to give a comparison of eight machine learning algorithms, including logistic regression (LR), random forest (RF), K-nearest neighbors (KNN), decision tree (DT), ada boost (AB), support vector machine (SVM), gradient boosting (GB), and Gaussian Naive Bayes (GNB) for breast cancer detection. The breast cancer Wisconsin (diagnostic) dataset is being utilized to validate the findings of this study. The comparison was made using the following performance metrics: accuracy, sensitivity, false omission rate, specificity, false discovery rate and area under curve. The LR method achieved a maximum accuracy of 99.12% among all eight algorithms and was compared to other comparable studies in the literature. The five features chosen are used to calculate the model's fidelity-to-interpretability ratio (FIR), which indicates how much interpretability was sacrificed for performance. The uniqueness of this work is the explainability approach taken in the model's performance, which aims to make the model's outputs more understandable and interpretable to healthcare experts.
Alzheimer's disease is largely the underlying cause of dementia due to its progressive neurodegen... more Alzheimer's disease is largely the underlying cause of dementia due to its progressive neurodegenerative nature among the elderly. The disease can be divided into five stages: Subjective Memory Concern (SMC), Mild Cognitive Impairment (MCI), Early MCI (EMCI), Late MCI (LMCI), and Alzheimer's Disease (AD). Alzheimer's disease is conventionally diagnosed using an MRI scan of the brain. In this research, we propose a fine-tuned convolutional neural network (CNN) classifier called AlzheimerNet, which can identify all five stages of Alzheimer's disease and the Normal Control (NC) class. The ADNI database's MRI scan dataset is obtained for use in training and testing the proposed model. To prepare the raw data for analysis, we applied the CLAHE image enhancement method. Data augmentation was used to remedy the unbalanced nature of the dataset and the resultant dataset consisted of 60000 image data on the 6 classes. Initially, five existing models including VGG16, MobileNetV2, AlexNet, ResNet50 and InceptionV3 were trained and tested to achieve test accuracies of 78.84%, 86.85%, 78.87%, 80.98% and 96.31% respectively. Since InceptionV3 provides the highest accuracy, this model is later modified to design the AlzheimerNet using RMSprop optimizer and learning rate 0.00001 to achieve the highest test accuracy of 98.67%. The five pre-trained models and the proposed fine-tuned model were compared in terms of various performance matrices to demonstrate whether the AlzheimerNet model is in fact performing better in classifying and detecting the six classes. An ablation study shows the hyperparameters used in the experiment. The suggested model outperforms the traditional methods for classifying Alzheimer's disease stages from brain MRI, as measured by a two-tailed Wilcoxon signed-rank test, with a significance of < 0.05. INDEX TERMS Alzheimer's disease, AlzheimerNet, deep learning, MRI imaging, multiclassification. The associate editor coordinating the review of this manuscript and approving it for publication was Mehul S. Raval .
algorithms, 2022
Early Prediction of Chronic Kidney Disease: A Comprehensive
Performance Analysis of Deep Learning... more Early Prediction of Chronic Kidney Disease: A Comprehensive
Performance Analysis of Deep Learning Models
Bulletin of Electrical Engineering and Informatics, 2022
Heart disease is among the leading causes for death globally. Thus, early identification and trea... more Heart disease is among the leading causes for death globally. Thus, early identification and treatment are indispensable to prevent the disease. In this work, we propose a framework based on machine learning algorithms to tackle such problems through the identification of risk variables associated to this disease. To ensure the success of our proposed model, influential data pre-processing and data transformation strategies are used to generate accurate data for the training model that utilizes the five most popular datasets (Hungarian, Stat log, Switzerland, Long Beach VA, and Cleveland) from UCI. The univariate feature selection technique is applied to identify essential features and during the training phase, classifiers, namely extreme gradient boosting (XGBoost), support vector machine (SVM), random forest (RF), gradient boosting (GB), and decision tree (DT), are deployed. Subsequently, various performance evaluations are measured to demonstrate accurate predictions using the introduced algorithms. The inclusion of Univariate results indicated that the DT classifier achieves a comparatively higher accuracy of around 97.75% than others. Thus, a machine learning approach is recognize, that can predict heart disease with high accuracy. Furthermore, the 10 attributes chosen are used to analyze the model's outcomes explainability, indicating which attributes are more significant in the model's outcome.
IEEE, 2022
In Bangladesh, the increased number of brokers has resulted in a price increase in the agricultur... more In Bangladesh, the increased number of brokers has resulted in a price increase in the agricultural goods market. They acquire items from farmers at a discount and resell them to consumers at a premium. This deprives farmers, while customers pay an unreasonable price for basic essentials. This study aims to develop an e-commerce mobile and web application called 'e-Farmers' Hut' to facilitate direct producer-to-customer engagement. The application goal to maximize the advantages to producers and customers by eliminating middlemen's control. Both applications operate sequentially on the same database. Farmers and customers each have their profile to which they may submit the relevant information. Additionally, to optimize facilities use, also introduced an electronic payment method. Customers may see lists of available items that verified farmers have posted. The application has passed the testing process with expected outcomes. The technology has the ability to considerably simplify direct sales and purchases of items between farmers and consumers.
IEEE, 2022
Liquefied petroleum gas (LPG) is commonly used for heating, cooking, automotive power, and variou... more Liquefied petroleum gas (LPG) is commonly used for heating, cooking, automotive power, and various other uses worldwide. LPG is a particularly flammable gas, and LPG leaks cause significant incidents. The cause may arise from improper installation to the use of faulty gas cylinders. Since LPG is an extremely volatile and flammable gas, a reliable safety system has been designed and developed using IoT (Internet of Things) capable of detecting gas leakage, turning on the emergency alarm, tracking the location, and sending alerts messages to users and nearest helpline number. The proposed model sort out into four modules, such as Gas Detection Module (GDM) always detect the gas leakage to avoid unexpected incidents; Location Detection Module (LDM) track the gas leakage location and pass the value to NM; Notification Module (NM) is responsible for generating the message service to notify the nearest help center and user; In the case of a gas leak, the Alarm Module (AM) is responsible for activating an emergency alarm. The result shows the system successfully performed. It can be noted that the proposed can be embedded with any environment, including home, office, ship, industry, etc.
IEEE, 2022
In recent years, people worldwide have been suffering from various types of kidney diseases indes... more In recent years, people worldwide have been suffering from various types of kidney diseases indescribably, among which chronic kidney disease (CKD) has exacerbated the situation. Early diagnosis of CKD is the only way to hinder the advancement of kidney disease in its initial stage. However, presently doctors can use machine learning classifier algorithms to identify the disease earlier than any other existing method. Here, this research work presents a method by using eight different machine learning (ML) algorithms that can promptly detect the infection of CKD considering the health condition dataset information of the patient. A dataset is used of nearly two months of that period delivered by the hospital to identify the plausibility of chronic kidney disease. This research study has used the Extra Tree Classifier (EXT), AdaBoost (ADB), K-Nearest Neighbors (KNN), Gradient Boosting (GB), Extreme Gradient Boosting (XGB), Decision Tree (DT), Gaussian Naïve Bayes (GNB) and Random Forest (RF) to obtain an optimum result of prediction. After pre-processing the data, this research work has applied the ML algorithms and compared their performances, and eventually, the precise outcome has been obtained. The performance is analyzed by using the F1-score, precision, accuracy, recall, and AUC score. According to the analysis results, K-Nearest Neighbors and Extra Tree Classifier have performed better than other algorithms for achieving an accuracy of 99% preceding the Gradient Boost, which stands at 98%.
International Journal of Advanced Computer Science and Applications, 2022
In the age of the industrial revolution, industry and machinery are elements of the utmost import... more In the age of the industrial revolution, industry and machinery are elements of the utmost importance to the development of human civilization. As industries are dependent on their machines, regular maintenance of these machines is required. However, if the machine is too big for humans to look after, we need a system that will observe these giants. This paper proposes a convolutional neural network-based system that detects faults in industrial machines by diagnosing motor sounds using accelerometers sensors. The sensors collect data from the machines and augment the data into 261756 samples to train (70%) and test (30%) the models for better accuracy. The sensor data are sent to the server through the wireless sensor network and decomposed using discrete wavelet transformation (DWT). This big data is processed to detect faults. The study shows that custom CNN architectures surpass the performance of the transfer learning-based MobileNetV2 fault diagnosis model. The system could successfully detect faults with up to 99.64% accuracy and 99.83% precision with the MobileNetV2 pretrained on the ImageNet Dataset. However, the Convolutional 1D and 2D architectures perform excellently with 100% accuracy and 100 % precision.
Journal of Personalized Medicine, 2022
In recent years, lung disease has increased manyfold, causing millions of casualties annually. To... more In recent years, lung disease has increased manyfold, causing millions of casualties annually. To combat the crisis, an efficient, reliable, and affordable lung disease diagnosis technique has become indispensable. In this study, a multiclass classification of lung disease from frontal chest X-ray imaging using a fine-tuned CNN model is proposed. The classification is conducted on 10 disease classes of the lungs, namely COVID-19, Effusion, Tuberculosis, Pneumonia, Lung Opacity, Mass, Nodule, Pneumothorax, and Pulmonary Fibrosis, along with the Normal class. The dataset is a collective dataset gathered from multiple sources. After pre-processing and balancing the dataset with eight augmentation techniques, a total of 80,000 X-ray images were fed to the model for classification purposes. Initially, eight pre-trained CNN models, AlexNet, GoogLeNet, InceptionV3, MobileNetV2, VGG16, ResNet 50, DenseNet121, and EfficientNetB7, were employed on the dataset. Among these, the VGG16 achieved the highest accuracy at 92.95%. To further improve the classification accuracy, LungNet22 was constructed upon the primary structure of the VGG16 model. An ablation study was used in the work to determine the different hyper-parameters. Using the Adam Optimizer, the proposed model achieved a commendable accuracy of 98.89%. To verify the performance of the model, several performance matrices, including the ROC curve and the AUC values, were computed as well.
Uploads
Papers by F M Javed Mehedi Shamrat
Performance Analysis of Deep Learning Models