The study evaluated several machine learning techniques for detecting phishing attacks, including... more The study evaluated several machine learning techniques for detecting phishing attacks, including Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Random Forest (RF), Decision Tree (DT), and Logistic Regression (LR). Two datasets were used-one from PhishTank and another from the UCI machine learning repository. Results showed that the Random Forest model achieved the highest accuracy across multiple metrics. On the PhishTank dataset, RF had the best K-fold cross-validation accuracy at 99.55%, feature selection accuracy at 99.00%, and hyperparameter tuning accuracy at 99.45%. The XGBoost model performed well too, with 99.16% K-fold accuracy on PhishTank. On the UCI dataset, XGBoost had the highest K-fold accuracy at 97.16%, while RF still demonstrated maximum accuracy for feature selection and hyperparameter tuning. Logistic Regression consistently showed the lowest accuracy across datasets and metrics. The proposed approach was validated against other researchers' work on PhishTank, achieving 98.80% accuracy, which was compared favorably. ROC curves further illustrated the strong performance, especially for the top-performing models. The study demonstrated that using selected features and hyperparameter tuning could enhance detection accuracy. The machine learning algorithms, particularly Random Forest, outperformed other state-ofthe-art techniques in accurately identifying phishing attacks. The high accuracy metrics indicate the proposed framework's effectiveness in detecting phishing attempts.
The widespread usage of internet-connected gadgets has led to a transformation in how individuals... more The widespread usage of internet-connected gadgets has led to a transformation in how individuals engage with technology, facilitating easy participation in various online activities such as social media, banking, and shopping. However, this proliferation has also provided an opportunity for fraudsters to exploit the Internet's anonymity through elaborate phishing schemes. These schemes aim to deceive users into disclosing personal information, including passwords and bank account details, often employing social engineering techniques. Consequently, the development of sophisticated phishing detection systems has become imperative to safeguard users' financial assets and digital identities. Many of these systems leverage state-of-the-art technology, with a significant reliance on machine learning methods for accurate and rapid detection of phishing attempts. Among these methods, deep learning algorithms have gained prominence due to their ability to efficiently process and analyze vast amounts of data. This paper presents a unique phishing detection system grounded in deep learning principles, employing a diverse array of techniques such as CNNs, attention networks, neural architects, and recurrent neural networks. The system's efficacy in real-time phishing attack detection is demonstrated through its focus on swift categorization of web pages based on URLs. Evaluation of the proposed method using a large dataset comprising nearly five million tagged URLs underscores its effectiveness.
This article is an open access article distributed under the terms and conditions of the Creative... more This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY
The purpose of this research is to propose a new method for identifying diabetic retinopathy usin... more The purpose of this research is to propose a new method for identifying diabetic retinopathy using retinal fundus images. Currently, identifying diabetic retinopathy from computerized fundus images is a challenging task in medical image processing and requires new strategies to be developed. The manual analysis of the retinal fundus is time-consuming and requires a significant amount of skill. To assist clinicians, this research develops a graphical user interface that integrates imaging algorithms to assess whether the patient's fundus image is affected by diabetic retinopathy. The diagnosis is made using a deep neural network, specifically the Resnet152-V2, which has been shown to have 100% accuracy in all evaluation criteria including accuracy, recall, precision, and F1 Score. The severity of the disease is displayed on the graphical user interface and the patient's information is stored in a local database. This proposed method can also be used by ophthalmologists as a backup option to support in disease detection, reducing the necessary processing time.
Diabetic retinopathy (DR) is a prevalent and potentially sight-threatening complication of diabet... more Diabetic retinopathy (DR) is a prevalent and potentially sight-threatening complication of diabetes mellitus. This condition ranks among the primary causes of blindness. Recently, we have seen an increase in interest in the use of advanced deep-learning methods in the analysis of medical images. Fundus photo with problems such as low contrast, noise, and uneven lighting. At the forefront of deep learning in image analysis are convolutional neural networks (CNNs), which show remarkable accuracy for distinguishing and categorizing severity levels of diabetic retinopathy in fundus images. Among these architectures, ResNet (Residual Network) CNN is effective. Validation of this approach is done through rigorous experimental evaluation and model training using established datasets such as EyePACS and MESSIDOR. These evaluations underline the potential of these models in the classification of diabetic retinopathy. The strategic alliance between medical expertise and cutting-edge technology is embodied by the integration of artificial intelligence models into cutting-edge devices, exemplified by smartphones. This integration facilitates real-time processing, leading to accelerated diagnosis and timely interventions. Such a combination of capabilities fits seamlessly into the Internet of Medical Things (IoMT) domain.
Precisely categorizing lung diseases is essential for effective medical treatments. This paper pr... more Precisely categorizing lung diseases is essential for effective medical treatments. This paper presents a comprehensive analysis of advanced methods in lung disease classification, with a focus on integrating diverse imaging techniques like computerized tomography (CT), X-rays, and magnetic resonance imaging (MRI). These imaging approaches collectively enhance the understanding of pulmonary conditions, aiding in early detection and differential diagnosis. The paper initially explains the fundamental principles of CT, MRI, and X-rays, highlighting their unique characteristics and roles in elucidating lung structures. It explores state-of-the-art methodologies, encompassing both traditional machine learning using engineered features and the expanding domain of deep learning utilizing neural networks to classify intricate diseases. A wide range of prevalent lung ailments, spanning from pneumonia and lung cancer to chronic obstructive pulmonary disease (COPD), are covered. Each domain delves into the considerations for adapting imaging modalities, involving data pre-processing, feature extraction, and algorithmic orchestration. Comparative evaluations of performance metrics offer insights into the effectiveness and limitations of each approach. Furthermore, the paper outlines the challenges associated with classifying lung diseases, including limited annotated data, complexities in model interpretation, and the seamless integration of algorithmic outcomes into clinical practices. As for future research avenues, the paper suggests innovative directions such as data augmentation, integrating multi-modal imaging information, and advancing transparent artificial intelligence (AI) frameworks to enhance their acceptance in clinical settings.
The COVID-19 influenza emerged and proved to be fatal, causing millions of deaths worldwide. Vacc... more The COVID-19 influenza emerged and proved to be fatal, causing millions of deaths worldwide. Vaccines were eventually discovered, effectively preventing the severe symptoms caused by the disease. However, some of the population (elderly and patients with comorbidities) are still vulnerable to severe symptoms such as breathlessness and chest pain. Identifying these patients in advance is imperative to prevent a bad prognosis. Hence, machine learning and deep learning algorithms have been used for early COVID-19 severity prediction using clinical and laboratory markers. The COVID-19 data was collected from two Manipal hospitals after obtaining ethical clearance. Multiple natureinspired feature selection algorithms are used to choose the most crucial markers. A maximum testing accuracy of 95% was achieved by the classifiers. The predictions obtained by the classifiers have been demystified using five explainable artificial intelligence techniques (XAI). According to XAI, the most important markers are c-reactive protein, basophils, lymphocytes, albumin, D-Dimer and neutrophils. The models could be deployed in various healthcare facilities to predict COVID-19 severity in advance so that appropriate treatments could be provided to mitigate a severe prognosis. The computer aided diagnostic method can also aid the healthcare professionals and ease the burden on already suffering healthcare infrastructure. The COVID-19 began in late 2019 and caused a significant uproar worldwide 1. Most patients experienced mildmoderate symptoms such as cough, cold, myalgia, sore throat, muscle pain, nausea, loss of taste/smell and headaches. However, people also developed severe symptoms such as accurate respiratory disorder syndrome (ARDS), severe hypoxia and multi-organ failure and succumbed to this deadly disease 2. As of today, the virus is still spreading, and new mutations are being created. Cytokine storm manifests in COVID-19 patients, distinguished by an enormous release of cytokines such as IL-6 and IL-1. This condition has led to the immune system attacking itself and has caused deaths in many Sars-Cov-2 patients 3. The severe symptoms of COVID-19 have decreased after the introduction of vaccines 4. However, some COVID-19 patients are still vulnerable to severe prognoses 3. Older patients and people with comorbidities such as hypertension, diabetes, cancer etc., are still at risk. It is crucial to identify these patients early so that appropriate medications and treatments can be provided to them to avoid unnecessary casualties. A few drugs have been created and shown to prevent the onset of severe COVID-19 symptoms 3. These medicines must be administered during the illness's initial stages to be effective. Artificial Intelligence (AI) applications have been extensively utilized in the healthcare sector 5-7. Diagnostic and prognostic models, decision support systems and predictive modelling are being developed to assist healthcare professionals using machine learning (ML). The above technologies are also being used in the fight against COVID-19 8-10. Explainable artificial intelligence (XAI) makes the models more transparent and understandable. The reasoning behind a patient prediction can be visually represented using XAI. It has also been utilized in various domains such as finance, engineering, pharmacy, medicine and commerce.
Early detection of lung disease is important for timely intervention and treatment, enhancing pat... more Early detection of lung disease is important for timely intervention and treatment, enhancing patient outcomes and decreasing healthcare cost. Chest X-rays are a widely employed imaging modality to examine the structures within the chest, including the lungs and surrounding tissues. Lung disease detection using chest X-rays is a critical application of medical imaging and artificial intelligence (AI) in healthcare. Recently, lung disease detection using deep learning (DL) becomes a significant research area, which has the potential to improve early detection rate and decrease mortality rate. Therefore, this article introduces a Multi-Feature Fusion Based Deep Transfer Learning with Enhanced Dung Beetle Optimization Algorithm (MFFTL-EDBOA) for lung disease detection and classification. The MFFTL-EDBOA technique aims to recognize the existence of lung diseases on CXR images. At the primary stage, the MFFTL-EDBOA technique uses adaptive filtering (AF) approach to remove the noise level. Besides, a multi-feature fusion-based feature extraction approach is developed based on three DL models namely DenseNet, EfficientNet, and MobileNet. For accurate lung disease detection and classification purposes, the convolutional fuzzy neural network (CFNN) approach is utilized. The hyperparameter tuning of the CFNN model occurs using the EDBOA. To illustrate the enhanced lung disease detection results of the MFFTL-EDBOA technique, a sequence of experiments is carried out on benchmark medical dataset from Kaggle repository. The experimental values highlighted the greater result of the MFFTL-EDBOA system over other recent approaches with maximum accuracy of 98.99%.
With fraud becoming more sophisticated, conventional detection methods are no longer effective, r... more With fraud becoming more sophisticated, conventional detection methods are no longer effective, resulting in a worldwide impact on customers and organizations. To tackle this, cutting-edge technologies like machine learning and blockchain are being utilized by several institutions. This article assesses the efficacy of XGBoost, KNN, CatBoost, and Random Forest in detecting real-time fraud during financial transactions. Additionally, the paper discusses how blockchain technology can create a secure and tamper-proof database for financial transactions used in fraud detection. Our proposed financial fraud detection approach was analyzed using the "Synthetic data from a financial payment system," revealing that 98.79% of the dataset comprised genuine transactions, while 1.212% were fraudulent. The results showed that CatBoost had the highest accuracy rate, exceeding 99.46%, while Random Forest had the lowest accuracy rate of 98.31% among all algorithms. A machine learning and blockchain technique has finally been proposed to identify fraudulent bank transactions.
Transfer learning may boost modeling speed. This research unifies the improved VGG16 model. Skipp... more Transfer learning may boost modeling speed. This research unifies the improved VGG16 model. Skipping VGG16's entirely connected layer and tying it to the layer following it improves CNN's architecture and reduces its processing burden. CNNs are classifying cat and dog photos. With a small dataset, CNN training may take time and resources. Transfer learning uses ImageNet models to handle similar challenges in new contexts. Transfer learning enhances CNN cat-dog classification. The proposed technique selects a suitable pre-trained model, freezes its layers to retain acquired properties, adds layers to learn task-specific features, and tunes it using the cat and dog dataset. Data augmentation improves model performance and prevents overfitting. Hyperparameter optimization and validation enhance model accuracy and speed. This method enhances cat and dog photo classification even without adequate data to train CNNs with an accuracy of 97%.
Email being an efficient, cost-effective, real-time communication mode results into effective pro... more Email being an efficient, cost-effective, real-time communication mode results into effective productivity among the professional in the organization. It constitutes almost 90% of daily office procedures in organizations, hence the productivity of organizations depends heavily on the text communicated in emails. The presented research work focuses on email profiling in organizations based on mail text interpretation and analysis. In the proposed work we will be working on datasets containing email communication of ENRON Corporation as test case. The profiling would be done using Text interpretation and analysis algorithm using machine learning algorithms. The BoW will be implemented to analyze and predict the characteristics of incoming and outgoing emails, then these could be mapped and profiled as per the behavior of employees into 3 categories of productive based on positive responses, neutral and non-productive based on negative responses.
Alzheimer's disease (AD) is a progressive neurodegenerative disease that affects cognition, behav... more Alzheimer's disease (AD) is a progressive neurodegenerative disease that affects cognition, behavior, and memory, eventually reaching a point where daily activities are impaired. Although there is currently no cure, initiating a well-considered management approach in the early stages can improve quality of life and potentially slow disease progression. Various machine learning (ML) techniques are widely used in clinical research to aid in the detection and tracking of disease states. Magnetic resonance imaging (MRI) is considered one of the most effective tools for diagnosing Alzheimer's disease. However, detecting subtle changes in the AD-affected brain in the early stages presents a significant challenge. The main challenges are the extremely small numbers of trained samples and larger feature descriptions. Hence, in this research, automatic AD can be diagnosed through the adoption of hybrid deep learning (DL) methodologies. For image pre-processing, Improved adaptive wiener filtering (IAWF) is utilized to enhance the acquired images. Then, the features are extracted by an effective hybrid method named Principal Component Analysis, which uses a Normalized Global Image Descriptor (PCA-NGIST) to extract the significant features from images without any image segmentation. Next, the best features are selected using the Improved Wild Horse Optimization algorithm (IWHO). Finally, the disease is diagnosed by hybrid Bi-directional Long Short-Term Memory with Artificial Neural Network (BiLSTM-ANN). The suggested method is implemented on the MATLAB platform. An accuracy of 99.22% is attained for the ADNI dataset and 98.96% for the OASIS dataset, which are comparatively better than the state-of-the-art methods.
A progressive neurodegenerative Alzheimer's Disease (AD) shrinks (atrophy) the brain and kills br... more A progressive neurodegenerative Alzheimer's Disease (AD) shrinks (atrophy) the brain and kills brain cells. On a global scale, Finland and the United Kingdom have the highest prevalence of AD, with 54.7 and 42.7 cases per 100,000 inhabitants respectively. In India, there are 14.60 AD cases for every 100,000 people and it is ranked 137 th in the world and 13 th in India among the 50 causes of death. At the beginning of the disease, a person may forget things, or have trouble remembering recent conversations. As the disease progresses, a person may experience severe memory loss and difficulty performing everyday tasks. Consequently, the only way to predict AD at an early stage is to use treatments that prevent the disease from progressing. This paper provides an overview of the current approaches to AD diagnosis and detection, with a focus on the use of Biomarkers and discussed the advantages and disadvantages of Machine Learning (ML) / Deep Learning (DL) techniques in early detection of AD. It also reviews the collection of AD datasets and the pre-processing techniques used in the studies. Although DL methods gives the best outcome measures such as accuracy, precision, F-measures, recall and Area Under the Curve (AUC) still DL techniques have some limitations that can be overcome via Quantum Computing (QC). QC is currently emerging new technology based on quantum theory where qubits are used to store the data instead of bits. It accomplishes 1000 times faster and also solves the complex problem within a fraction of time than the classical computers.
The new coronavirus, or COVID-19, is a severe respiratory infection caused by the SARS-CoV-2 viru... more The new coronavirus, or COVID-19, is a severe respiratory infection caused by the SARS-CoV-2 virus. The pandemic began in Wuhan, China in late 2019 and quickly spread globally, resulting in widespread illness, death, and major disruptions to economies and societies. The virus mainly spreads via respiratory droplets, and typical symptoms include fatigue, fever, coughing, and shortness of breath. While some individuals may have mild symptoms, others can develop severe illness and potentially fatal complications. The COVID-19 pandemic has prompted widespread efforts to slow it's spread, including lockdowns, social distancing measures, and widespread use of personal protective equipment. Efforts to mitigate the impact of COVID-19 are ongoing, including research into treatments and vaccines. It becomes even more crucial in countries where laboratory kits is unavailable for testing. So, to diagnose COVID-19 disease, we tested a model built upon image processing as well as deep learning approaches. A COVID-19 infection could proceed to pneumonia, which could be diagnosed and treated with a chest X-ray. Here, we can create a dataset with 1403 chest Xrays. The dataset is loaded when the chest CT scans from COVID-19 have been acquired. We first propose pre-processing methods, use Image Augmentation, then make use of CT scans to determine whether or not COVID-19 (chest x-ray images) is positive or negative. Using deep learning to interpret images implementations are employed to achieve this.
With the development of Internet technology, online learning is becoming more and more popular. H... more With the development of Internet technology, online learning is becoming more and more popular. However, college students have different online learning behaviors and attitudes toward artificial intelligence (AI) learning tools. In this paper, a portrait model is proposed for college students, which focuses on their online learning behavior and attitudes toward AI learning tools. Moreover, the proposed portrait model is built based on AI technology, i.e., random forest algorithm and long short-term memory (LSTM) algorithm are applied. In this model, there are three main parts: data pre-processing, building a multidimensional label system, and portrait model of college students. Firstly, the information collected through the questionnaire is quantified and its quality is improved by deleting invalid data. Then, a multi-dimensional label system is built for college students' portraits, including basic attributes, online behavior attributes, behavioral attributes of using learning software, and psychological attributes of AI learning tools. Since each label consists of multiple indexes, the variance-based filtering method is used to streamline the indexes of online behavior attributes and behavioral attributes of using learning software, the random forest algorithm is applied to reduce the dimension of psychological attributes of AI learning tools. Next, the portrait of college students' online learning behavior is realized by the K-means clustering algorithm, and LSTM algorithm is performed to get the mapping mechanism between data and portrait categories. Through the mapping mechanism, the portrait of any college student can be obtained quickly. Finally, the validity of the proposed model is verified by analyzing the questionnaire results of college students. Additionally, the portrait results provide a data basis for the development and popularization of AI learning tools.
Alzheimer's disease (AD) is a progressive neurodegenerative disease mostly seen in the elderly af... more Alzheimer's disease (AD) is a progressive neurodegenerative disease mostly seen in the elderly affecting memory, thinking, and the ability to perform daily activities. As the incidence of AD is expected to grow in the coming years, early detection is crucial to handle the disease. At present, the primary imaging modality to diagnose an AD is based on cranial magnetic resonance imaging (MRI). The study aims to investigate the diagnostic capacity of convolutional neural network models via transfer learning in the recognition of brain atrophy and neuronal loss depicting AD from cranial MRI scans. Different neural configurations were tested to determine the top-performing model. The top performing model was obtained by ResNet50 with 0.001 learning rate yielding the highest AUC at 93.5, 85% accuracy, 88% recall, 82% specificity, 80% precision and 84% F1score. The best model was also incorporated to a simple application developed to predict AD. The early and immediate recognition of AD with transfer learning methods can cascade prompt referral to a neurologist or a geriatric medicine physician as well as the institution of appropriate therapeutic measures to mitigate the adverse impact of brain atrophy and to improve the quality of life of AD patients.
Fast and efficient landslide detection plays an important role in post-disaster rescue and risk a... more Fast and efficient landslide detection plays an important role in post-disaster rescue and risk assessment. Existing convolution neural network (CNN) based landslide detection methods are difficult to exploit global long-distance dependencies due to limited receptive fields. Considering that landslide occurrence is susceptible to local and global conditions, we propose a novel multi-scale feature fusion scene parsing (MFFSP) framework to explore information at different scales by coupling CNN with Transformer to learn local and global clues for landslide detection based on satellite data. In the encoder, we design three modules, visual geometry module (VGM), residual learning module (RLM), and Transformer module (TRM) to exploit multi-scale features. Specifically, VGM and RLM are constructed based on convolution operations to explore local features by learning low-level and middle-level information, while TRM is built based on self-attention mechanism to learn long-distance dependencies. In the decoder, TRM and VGM are further extended to motivate the model to mine long-distance dependencies and detailed spatial information by deeply fusing features from multiple scales. To demonstrate the performance of the model, we employ two study areas with four test regions to conduct experiments and compare with seven state-of-the-art deep learning models. Extensive experiments demonstrate that MFFSP greatly outperforms other algorithms. In addition, we conduct numerous ablation experiments, proving that MFFSP fully combines the complementary advantages of CNN and Transformer to mine robust features.
The study evaluated several machine learning techniques for detecting phishing attacks, including... more The study evaluated several machine learning techniques for detecting phishing attacks, including Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Random Forest (RF), Decision Tree (DT), and Logistic Regression (LR). Two datasets were used-one from PhishTank and another from the UCI machine learning repository. Results showed that the Random Forest model achieved the highest accuracy across multiple metrics. On the PhishTank dataset, RF had the best K-fold cross-validation accuracy at 99.55%, feature selection accuracy at 99.00%, and hyperparameter tuning accuracy at 99.45%. The XGBoost model performed well too, with 99.16% K-fold accuracy on PhishTank. On the UCI dataset, XGBoost had the highest K-fold accuracy at 97.16%, while RF still demonstrated maximum accuracy for feature selection and hyperparameter tuning. Logistic Regression consistently showed the lowest accuracy across datasets and metrics. The proposed approach was validated against other researchers' work on PhishTank, achieving 98.80% accuracy, which was compared favorably. ROC curves further illustrated the strong performance, especially for the top-performing models. The study demonstrated that using selected features and hyperparameter tuning could enhance detection accuracy. The machine learning algorithms, particularly Random Forest, outperformed other state-ofthe-art techniques in accurately identifying phishing attacks. The high accuracy metrics indicate the proposed framework's effectiveness in detecting phishing attempts.
The widespread usage of internet-connected gadgets has led to a transformation in how individuals... more The widespread usage of internet-connected gadgets has led to a transformation in how individuals engage with technology, facilitating easy participation in various online activities such as social media, banking, and shopping. However, this proliferation has also provided an opportunity for fraudsters to exploit the Internet's anonymity through elaborate phishing schemes. These schemes aim to deceive users into disclosing personal information, including passwords and bank account details, often employing social engineering techniques. Consequently, the development of sophisticated phishing detection systems has become imperative to safeguard users' financial assets and digital identities. Many of these systems leverage state-of-the-art technology, with a significant reliance on machine learning methods for accurate and rapid detection of phishing attempts. Among these methods, deep learning algorithms have gained prominence due to their ability to efficiently process and analyze vast amounts of data. This paper presents a unique phishing detection system grounded in deep learning principles, employing a diverse array of techniques such as CNNs, attention networks, neural architects, and recurrent neural networks. The system's efficacy in real-time phishing attack detection is demonstrated through its focus on swift categorization of web pages based on URLs. Evaluation of the proposed method using a large dataset comprising nearly five million tagged URLs underscores its effectiveness.
This article is an open access article distributed under the terms and conditions of the Creative... more This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY
The purpose of this research is to propose a new method for identifying diabetic retinopathy usin... more The purpose of this research is to propose a new method for identifying diabetic retinopathy using retinal fundus images. Currently, identifying diabetic retinopathy from computerized fundus images is a challenging task in medical image processing and requires new strategies to be developed. The manual analysis of the retinal fundus is time-consuming and requires a significant amount of skill. To assist clinicians, this research develops a graphical user interface that integrates imaging algorithms to assess whether the patient's fundus image is affected by diabetic retinopathy. The diagnosis is made using a deep neural network, specifically the Resnet152-V2, which has been shown to have 100% accuracy in all evaluation criteria including accuracy, recall, precision, and F1 Score. The severity of the disease is displayed on the graphical user interface and the patient's information is stored in a local database. This proposed method can also be used by ophthalmologists as a backup option to support in disease detection, reducing the necessary processing time.
Diabetic retinopathy (DR) is a prevalent and potentially sight-threatening complication of diabet... more Diabetic retinopathy (DR) is a prevalent and potentially sight-threatening complication of diabetes mellitus. This condition ranks among the primary causes of blindness. Recently, we have seen an increase in interest in the use of advanced deep-learning methods in the analysis of medical images. Fundus photo with problems such as low contrast, noise, and uneven lighting. At the forefront of deep learning in image analysis are convolutional neural networks (CNNs), which show remarkable accuracy for distinguishing and categorizing severity levels of diabetic retinopathy in fundus images. Among these architectures, ResNet (Residual Network) CNN is effective. Validation of this approach is done through rigorous experimental evaluation and model training using established datasets such as EyePACS and MESSIDOR. These evaluations underline the potential of these models in the classification of diabetic retinopathy. The strategic alliance between medical expertise and cutting-edge technology is embodied by the integration of artificial intelligence models into cutting-edge devices, exemplified by smartphones. This integration facilitates real-time processing, leading to accelerated diagnosis and timely interventions. Such a combination of capabilities fits seamlessly into the Internet of Medical Things (IoMT) domain.
Precisely categorizing lung diseases is essential for effective medical treatments. This paper pr... more Precisely categorizing lung diseases is essential for effective medical treatments. This paper presents a comprehensive analysis of advanced methods in lung disease classification, with a focus on integrating diverse imaging techniques like computerized tomography (CT), X-rays, and magnetic resonance imaging (MRI). These imaging approaches collectively enhance the understanding of pulmonary conditions, aiding in early detection and differential diagnosis. The paper initially explains the fundamental principles of CT, MRI, and X-rays, highlighting their unique characteristics and roles in elucidating lung structures. It explores state-of-the-art methodologies, encompassing both traditional machine learning using engineered features and the expanding domain of deep learning utilizing neural networks to classify intricate diseases. A wide range of prevalent lung ailments, spanning from pneumonia and lung cancer to chronic obstructive pulmonary disease (COPD), are covered. Each domain delves into the considerations for adapting imaging modalities, involving data pre-processing, feature extraction, and algorithmic orchestration. Comparative evaluations of performance metrics offer insights into the effectiveness and limitations of each approach. Furthermore, the paper outlines the challenges associated with classifying lung diseases, including limited annotated data, complexities in model interpretation, and the seamless integration of algorithmic outcomes into clinical practices. As for future research avenues, the paper suggests innovative directions such as data augmentation, integrating multi-modal imaging information, and advancing transparent artificial intelligence (AI) frameworks to enhance their acceptance in clinical settings.
The COVID-19 influenza emerged and proved to be fatal, causing millions of deaths worldwide. Vacc... more The COVID-19 influenza emerged and proved to be fatal, causing millions of deaths worldwide. Vaccines were eventually discovered, effectively preventing the severe symptoms caused by the disease. However, some of the population (elderly and patients with comorbidities) are still vulnerable to severe symptoms such as breathlessness and chest pain. Identifying these patients in advance is imperative to prevent a bad prognosis. Hence, machine learning and deep learning algorithms have been used for early COVID-19 severity prediction using clinical and laboratory markers. The COVID-19 data was collected from two Manipal hospitals after obtaining ethical clearance. Multiple natureinspired feature selection algorithms are used to choose the most crucial markers. A maximum testing accuracy of 95% was achieved by the classifiers. The predictions obtained by the classifiers have been demystified using five explainable artificial intelligence techniques (XAI). According to XAI, the most important markers are c-reactive protein, basophils, lymphocytes, albumin, D-Dimer and neutrophils. The models could be deployed in various healthcare facilities to predict COVID-19 severity in advance so that appropriate treatments could be provided to mitigate a severe prognosis. The computer aided diagnostic method can also aid the healthcare professionals and ease the burden on already suffering healthcare infrastructure. The COVID-19 began in late 2019 and caused a significant uproar worldwide 1. Most patients experienced mildmoderate symptoms such as cough, cold, myalgia, sore throat, muscle pain, nausea, loss of taste/smell and headaches. However, people also developed severe symptoms such as accurate respiratory disorder syndrome (ARDS), severe hypoxia and multi-organ failure and succumbed to this deadly disease 2. As of today, the virus is still spreading, and new mutations are being created. Cytokine storm manifests in COVID-19 patients, distinguished by an enormous release of cytokines such as IL-6 and IL-1. This condition has led to the immune system attacking itself and has caused deaths in many Sars-Cov-2 patients 3. The severe symptoms of COVID-19 have decreased after the introduction of vaccines 4. However, some COVID-19 patients are still vulnerable to severe prognoses 3. Older patients and people with comorbidities such as hypertension, diabetes, cancer etc., are still at risk. It is crucial to identify these patients early so that appropriate medications and treatments can be provided to them to avoid unnecessary casualties. A few drugs have been created and shown to prevent the onset of severe COVID-19 symptoms 3. These medicines must be administered during the illness's initial stages to be effective. Artificial Intelligence (AI) applications have been extensively utilized in the healthcare sector 5-7. Diagnostic and prognostic models, decision support systems and predictive modelling are being developed to assist healthcare professionals using machine learning (ML). The above technologies are also being used in the fight against COVID-19 8-10. Explainable artificial intelligence (XAI) makes the models more transparent and understandable. The reasoning behind a patient prediction can be visually represented using XAI. It has also been utilized in various domains such as finance, engineering, pharmacy, medicine and commerce.
Early detection of lung disease is important for timely intervention and treatment, enhancing pat... more Early detection of lung disease is important for timely intervention and treatment, enhancing patient outcomes and decreasing healthcare cost. Chest X-rays are a widely employed imaging modality to examine the structures within the chest, including the lungs and surrounding tissues. Lung disease detection using chest X-rays is a critical application of medical imaging and artificial intelligence (AI) in healthcare. Recently, lung disease detection using deep learning (DL) becomes a significant research area, which has the potential to improve early detection rate and decrease mortality rate. Therefore, this article introduces a Multi-Feature Fusion Based Deep Transfer Learning with Enhanced Dung Beetle Optimization Algorithm (MFFTL-EDBOA) for lung disease detection and classification. The MFFTL-EDBOA technique aims to recognize the existence of lung diseases on CXR images. At the primary stage, the MFFTL-EDBOA technique uses adaptive filtering (AF) approach to remove the noise level. Besides, a multi-feature fusion-based feature extraction approach is developed based on three DL models namely DenseNet, EfficientNet, and MobileNet. For accurate lung disease detection and classification purposes, the convolutional fuzzy neural network (CFNN) approach is utilized. The hyperparameter tuning of the CFNN model occurs using the EDBOA. To illustrate the enhanced lung disease detection results of the MFFTL-EDBOA technique, a sequence of experiments is carried out on benchmark medical dataset from Kaggle repository. The experimental values highlighted the greater result of the MFFTL-EDBOA system over other recent approaches with maximum accuracy of 98.99%.
With fraud becoming more sophisticated, conventional detection methods are no longer effective, r... more With fraud becoming more sophisticated, conventional detection methods are no longer effective, resulting in a worldwide impact on customers and organizations. To tackle this, cutting-edge technologies like machine learning and blockchain are being utilized by several institutions. This article assesses the efficacy of XGBoost, KNN, CatBoost, and Random Forest in detecting real-time fraud during financial transactions. Additionally, the paper discusses how blockchain technology can create a secure and tamper-proof database for financial transactions used in fraud detection. Our proposed financial fraud detection approach was analyzed using the "Synthetic data from a financial payment system," revealing that 98.79% of the dataset comprised genuine transactions, while 1.212% were fraudulent. The results showed that CatBoost had the highest accuracy rate, exceeding 99.46%, while Random Forest had the lowest accuracy rate of 98.31% among all algorithms. A machine learning and blockchain technique has finally been proposed to identify fraudulent bank transactions.
Transfer learning may boost modeling speed. This research unifies the improved VGG16 model. Skipp... more Transfer learning may boost modeling speed. This research unifies the improved VGG16 model. Skipping VGG16's entirely connected layer and tying it to the layer following it improves CNN's architecture and reduces its processing burden. CNNs are classifying cat and dog photos. With a small dataset, CNN training may take time and resources. Transfer learning uses ImageNet models to handle similar challenges in new contexts. Transfer learning enhances CNN cat-dog classification. The proposed technique selects a suitable pre-trained model, freezes its layers to retain acquired properties, adds layers to learn task-specific features, and tunes it using the cat and dog dataset. Data augmentation improves model performance and prevents overfitting. Hyperparameter optimization and validation enhance model accuracy and speed. This method enhances cat and dog photo classification even without adequate data to train CNNs with an accuracy of 97%.
Email being an efficient, cost-effective, real-time communication mode results into effective pro... more Email being an efficient, cost-effective, real-time communication mode results into effective productivity among the professional in the organization. It constitutes almost 90% of daily office procedures in organizations, hence the productivity of organizations depends heavily on the text communicated in emails. The presented research work focuses on email profiling in organizations based on mail text interpretation and analysis. In the proposed work we will be working on datasets containing email communication of ENRON Corporation as test case. The profiling would be done using Text interpretation and analysis algorithm using machine learning algorithms. The BoW will be implemented to analyze and predict the characteristics of incoming and outgoing emails, then these could be mapped and profiled as per the behavior of employees into 3 categories of productive based on positive responses, neutral and non-productive based on negative responses.
Alzheimer's disease (AD) is a progressive neurodegenerative disease that affects cognition, behav... more Alzheimer's disease (AD) is a progressive neurodegenerative disease that affects cognition, behavior, and memory, eventually reaching a point where daily activities are impaired. Although there is currently no cure, initiating a well-considered management approach in the early stages can improve quality of life and potentially slow disease progression. Various machine learning (ML) techniques are widely used in clinical research to aid in the detection and tracking of disease states. Magnetic resonance imaging (MRI) is considered one of the most effective tools for diagnosing Alzheimer's disease. However, detecting subtle changes in the AD-affected brain in the early stages presents a significant challenge. The main challenges are the extremely small numbers of trained samples and larger feature descriptions. Hence, in this research, automatic AD can be diagnosed through the adoption of hybrid deep learning (DL) methodologies. For image pre-processing, Improved adaptive wiener filtering (IAWF) is utilized to enhance the acquired images. Then, the features are extracted by an effective hybrid method named Principal Component Analysis, which uses a Normalized Global Image Descriptor (PCA-NGIST) to extract the significant features from images without any image segmentation. Next, the best features are selected using the Improved Wild Horse Optimization algorithm (IWHO). Finally, the disease is diagnosed by hybrid Bi-directional Long Short-Term Memory with Artificial Neural Network (BiLSTM-ANN). The suggested method is implemented on the MATLAB platform. An accuracy of 99.22% is attained for the ADNI dataset and 98.96% for the OASIS dataset, which are comparatively better than the state-of-the-art methods.
A progressive neurodegenerative Alzheimer's Disease (AD) shrinks (atrophy) the brain and kills br... more A progressive neurodegenerative Alzheimer's Disease (AD) shrinks (atrophy) the brain and kills brain cells. On a global scale, Finland and the United Kingdom have the highest prevalence of AD, with 54.7 and 42.7 cases per 100,000 inhabitants respectively. In India, there are 14.60 AD cases for every 100,000 people and it is ranked 137 th in the world and 13 th in India among the 50 causes of death. At the beginning of the disease, a person may forget things, or have trouble remembering recent conversations. As the disease progresses, a person may experience severe memory loss and difficulty performing everyday tasks. Consequently, the only way to predict AD at an early stage is to use treatments that prevent the disease from progressing. This paper provides an overview of the current approaches to AD diagnosis and detection, with a focus on the use of Biomarkers and discussed the advantages and disadvantages of Machine Learning (ML) / Deep Learning (DL) techniques in early detection of AD. It also reviews the collection of AD datasets and the pre-processing techniques used in the studies. Although DL methods gives the best outcome measures such as accuracy, precision, F-measures, recall and Area Under the Curve (AUC) still DL techniques have some limitations that can be overcome via Quantum Computing (QC). QC is currently emerging new technology based on quantum theory where qubits are used to store the data instead of bits. It accomplishes 1000 times faster and also solves the complex problem within a fraction of time than the classical computers.
The new coronavirus, or COVID-19, is a severe respiratory infection caused by the SARS-CoV-2 viru... more The new coronavirus, or COVID-19, is a severe respiratory infection caused by the SARS-CoV-2 virus. The pandemic began in Wuhan, China in late 2019 and quickly spread globally, resulting in widespread illness, death, and major disruptions to economies and societies. The virus mainly spreads via respiratory droplets, and typical symptoms include fatigue, fever, coughing, and shortness of breath. While some individuals may have mild symptoms, others can develop severe illness and potentially fatal complications. The COVID-19 pandemic has prompted widespread efforts to slow it's spread, including lockdowns, social distancing measures, and widespread use of personal protective equipment. Efforts to mitigate the impact of COVID-19 are ongoing, including research into treatments and vaccines. It becomes even more crucial in countries where laboratory kits is unavailable for testing. So, to diagnose COVID-19 disease, we tested a model built upon image processing as well as deep learning approaches. A COVID-19 infection could proceed to pneumonia, which could be diagnosed and treated with a chest X-ray. Here, we can create a dataset with 1403 chest Xrays. The dataset is loaded when the chest CT scans from COVID-19 have been acquired. We first propose pre-processing methods, use Image Augmentation, then make use of CT scans to determine whether or not COVID-19 (chest x-ray images) is positive or negative. Using deep learning to interpret images implementations are employed to achieve this.
With the development of Internet technology, online learning is becoming more and more popular. H... more With the development of Internet technology, online learning is becoming more and more popular. However, college students have different online learning behaviors and attitudes toward artificial intelligence (AI) learning tools. In this paper, a portrait model is proposed for college students, which focuses on their online learning behavior and attitudes toward AI learning tools. Moreover, the proposed portrait model is built based on AI technology, i.e., random forest algorithm and long short-term memory (LSTM) algorithm are applied. In this model, there are three main parts: data pre-processing, building a multidimensional label system, and portrait model of college students. Firstly, the information collected through the questionnaire is quantified and its quality is improved by deleting invalid data. Then, a multi-dimensional label system is built for college students' portraits, including basic attributes, online behavior attributes, behavioral attributes of using learning software, and psychological attributes of AI learning tools. Since each label consists of multiple indexes, the variance-based filtering method is used to streamline the indexes of online behavior attributes and behavioral attributes of using learning software, the random forest algorithm is applied to reduce the dimension of psychological attributes of AI learning tools. Next, the portrait of college students' online learning behavior is realized by the K-means clustering algorithm, and LSTM algorithm is performed to get the mapping mechanism between data and portrait categories. Through the mapping mechanism, the portrait of any college student can be obtained quickly. Finally, the validity of the proposed model is verified by analyzing the questionnaire results of college students. Additionally, the portrait results provide a data basis for the development and popularization of AI learning tools.
Alzheimer's disease (AD) is a progressive neurodegenerative disease mostly seen in the elderly af... more Alzheimer's disease (AD) is a progressive neurodegenerative disease mostly seen in the elderly affecting memory, thinking, and the ability to perform daily activities. As the incidence of AD is expected to grow in the coming years, early detection is crucial to handle the disease. At present, the primary imaging modality to diagnose an AD is based on cranial magnetic resonance imaging (MRI). The study aims to investigate the diagnostic capacity of convolutional neural network models via transfer learning in the recognition of brain atrophy and neuronal loss depicting AD from cranial MRI scans. Different neural configurations were tested to determine the top-performing model. The top performing model was obtained by ResNet50 with 0.001 learning rate yielding the highest AUC at 93.5, 85% accuracy, 88% recall, 82% specificity, 80% precision and 84% F1score. The best model was also incorporated to a simple application developed to predict AD. The early and immediate recognition of AD with transfer learning methods can cascade prompt referral to a neurologist or a geriatric medicine physician as well as the institution of appropriate therapeutic measures to mitigate the adverse impact of brain atrophy and to improve the quality of life of AD patients.
Fast and efficient landslide detection plays an important role in post-disaster rescue and risk a... more Fast and efficient landslide detection plays an important role in post-disaster rescue and risk assessment. Existing convolution neural network (CNN) based landslide detection methods are difficult to exploit global long-distance dependencies due to limited receptive fields. Considering that landslide occurrence is susceptible to local and global conditions, we propose a novel multi-scale feature fusion scene parsing (MFFSP) framework to explore information at different scales by coupling CNN with Transformer to learn local and global clues for landslide detection based on satellite data. In the encoder, we design three modules, visual geometry module (VGM), residual learning module (RLM), and Transformer module (TRM) to exploit multi-scale features. Specifically, VGM and RLM are constructed based on convolution operations to explore local features by learning low-level and middle-level information, while TRM is built based on self-attention mechanism to learn long-distance dependencies. In the decoder, TRM and VGM are further extended to motivate the model to mine long-distance dependencies and detailed spatial information by deeply fusing features from multiple scales. To demonstrate the performance of the model, we employ two study areas with four test regions to conduct experiments and compare with seven state-of-the-art deep learning models. Extensive experiments demonstrate that MFFSP greatly outperforms other algorithms. In addition, we conduct numerous ablation experiments, proving that MFFSP fully combines the complementary advantages of CNN and Transformer to mine robust features.
Uploads
Papers by Joyece Jane