The prompt spread of COVID-19 has emphasized the necessity for effective and precise diagnostic tools.In this article,a hybrid approach in terms of datasets as well as the methodology by utilizing a previously unexplo...The prompt spread of COVID-19 has emphasized the necessity for effective and precise diagnostic tools.In this article,a hybrid approach in terms of datasets as well as the methodology by utilizing a previously unexplored dataset obtained from a private hospital for detecting COVID-19,pneumonia,and normal conditions in chest X-ray images(CXIs)is proposed coupled with Explainable Artificial Intelligence(XAI).Our study leverages less preprocessing with pre-trained cutting-edge models like InceptionV3,VGG16,and VGG19 that excel in the task of feature extraction.The methodology is further enhanced by the inclusion of the t-SNE(t-Distributed Stochastic Neighbor Embedding)technique for visualizing the extracted image features and Contrast Limited Adaptive Histogram Equalization(CLAHE)to improve images before extraction of features.Additionally,an AttentionMechanism is utilized,which helps clarify how the modelmakes decisions,which builds trust in artificial intelligence(AI)systems.To evaluate the effectiveness of the proposed approach,both benchmark datasets and a private dataset obtained with permissions from Jinnah PostgraduateMedical Center(JPMC)in Karachi,Pakistan,are utilized.In 12 experiments,VGG19 showcased remarkable performance in the hybrid dataset approach,achieving 100%accuracy in COVID-19 vs.pneumonia classification and 97%in distinguishing normal cases.Overall,across all classes,the approach achieved 98%accuracy,demonstrating its efficiency in detecting COVID-19 and differentiating it fromother chest disorders(Pneumonia and healthy)while also providing insights into the decision-making process of the models.展开更多
Type 2 diabetes patients often suffer from microvascular compli-cations of diabetes.These complications,in turn,often lead to vision impair-ment.Diabetic Retinopathy(DR)detection in its early stage can rescue people f...Type 2 diabetes patients often suffer from microvascular compli-cations of diabetes.These complications,in turn,often lead to vision impair-ment.Diabetic Retinopathy(DR)detection in its early stage can rescue people from long-term complications that could lead to permanent blindness.In this study,we propose a complex deep convolutional neural network architecture with an inception module for automated diagnosis of DR.The proposed novel Hybrid Inception U-Net(HIUNET)comprises various inception modules connected in the U-Net fashion using activation maximization and filter map to produce the image mask.First,inception blocks were used to enlarge the model’s width by substituting it with primary convolutional layers.Then,aggregation blocks were used to deepen the model to extract more compact and discriminating features.Finally,the downsampling blocks were adopted to reduce the feature map size to decrease the learning time,and the upsam-pling blocks were used to resize the feature maps.This methodology ensured high prominence to lesion regions compared to the non-lesion regions.The performance of the proposed model was assessed on two benchmark compet-itive datasets called Asia Pacific Tele-Ophthalmology Society(APTOS)and KAGGLE,attaining accuracy rates of 95%and 92%,respectively.展开更多
基金National Natural Science Foundation of China(No.62075177)the Open Research Fund of CAS Key Laboratory of Spectral Imaging Technology(No.LSIT202005W)+1 种基金the 111 Project(No.B17035)。
基金the Deanship of Graduate Studies and Scientific Research at Qassim University for financial support(QU-APC-2024-9/1).
文摘The prompt spread of COVID-19 has emphasized the necessity for effective and precise diagnostic tools.In this article,a hybrid approach in terms of datasets as well as the methodology by utilizing a previously unexplored dataset obtained from a private hospital for detecting COVID-19,pneumonia,and normal conditions in chest X-ray images(CXIs)is proposed coupled with Explainable Artificial Intelligence(XAI).Our study leverages less preprocessing with pre-trained cutting-edge models like InceptionV3,VGG16,and VGG19 that excel in the task of feature extraction.The methodology is further enhanced by the inclusion of the t-SNE(t-Distributed Stochastic Neighbor Embedding)technique for visualizing the extracted image features and Contrast Limited Adaptive Histogram Equalization(CLAHE)to improve images before extraction of features.Additionally,an AttentionMechanism is utilized,which helps clarify how the modelmakes decisions,which builds trust in artificial intelligence(AI)systems.To evaluate the effectiveness of the proposed approach,both benchmark datasets and a private dataset obtained with permissions from Jinnah PostgraduateMedical Center(JPMC)in Karachi,Pakistan,are utilized.In 12 experiments,VGG19 showcased remarkable performance in the hybrid dataset approach,achieving 100%accuracy in COVID-19 vs.pneumonia classification and 97%in distinguishing normal cases.Overall,across all classes,the approach achieved 98%accuracy,demonstrating its efficiency in detecting COVID-19 and differentiating it fromother chest disorders(Pneumonia and healthy)while also providing insights into the decision-making process of the models.
文摘Type 2 diabetes patients often suffer from microvascular compli-cations of diabetes.These complications,in turn,often lead to vision impair-ment.Diabetic Retinopathy(DR)detection in its early stage can rescue people from long-term complications that could lead to permanent blindness.In this study,we propose a complex deep convolutional neural network architecture with an inception module for automated diagnosis of DR.The proposed novel Hybrid Inception U-Net(HIUNET)comprises various inception modules connected in the U-Net fashion using activation maximization and filter map to produce the image mask.First,inception blocks were used to enlarge the model’s width by substituting it with primary convolutional layers.Then,aggregation blocks were used to deepen the model to extract more compact and discriminating features.Finally,the downsampling blocks were adopted to reduce the feature map size to decrease the learning time,and the upsam-pling blocks were used to resize the feature maps.This methodology ensured high prominence to lesion regions compared to the non-lesion regions.The performance of the proposed model was assessed on two benchmark compet-itive datasets called Asia Pacific Tele-Ophthalmology Society(APTOS)and KAGGLE,attaining accuracy rates of 95%and 92%,respectively.
