ABSTRACT
BACKGROUND: Although great advances in artificial intelligence for interpreting small bowel capsule endoscopy (SBCE) images have been made in recent years, its practical use is still limited. The aim of this study was to develop a more practical convolutional neural network (CNN) algorithm for the automatic detection of various small bowel lesions. METHODS: A total of 7556 images were collected for the training dataset from 526 SBCE videos. Abnormal images were classified into two categories: hemorrhagic lesions (red spot/angioectasia/active bleeding) and ulcerative lesions (erosion/ulcer/stricture). A CNN algorithm based on VGGNet was trained in two different ways: the combined model (hemorrhagic and ulcerative lesions trained separately) and the binary model (all abnormal images trained without discrimination). The detected lesions were visualized using a gradient class activation map (Grad-CAM). The two models were validated using 5,760 independent images taken at two other academic hospitals. RESULTS: Both the combined and binary models acquired high accuracy for lesion detection, and the difference between the two models was not significant (96.83% vs 96.62%, P = 0.122). However, the combined model showed higher sensitivity (97.61% vs 95.07%, P < 0.001) and higher accuracy for individual lesions from the hemorrhagic and ulcerative categories than the binary model. The combined model also revealed more accurate localization of the culprit area on images evaluated by the Grad-CAM. CONCLUSIONS: Diagnostic sensitivity and classification of small bowel lesions using a convolutional neural network are improved by the independent training for hemorrhagic and ulcerative lesions. Grad-CAM is highly effective in localizing the lesions.
Subject(s)
Capsule Endoscopy , Deep Learning , Artificial Intelligence , Humans , Intestine, Small/diagnostic imaging , Neural Networks, ComputerABSTRACT
This study presents an up-to-date survey of the use of artificial intelligence (AI) in the field of otorhinolaryngology, considering opportunities, research challenges, and research directions. We searched PubMed, the Cochrane Central Register of Controlled Trials, Embase, and the Web of Science. We initially retrieved 458 articles. The exclusion of non-English publications and duplicates yielded a total of 90 remaining studies. These 90 studies were divided into those analyzing medical images, voice, medical devices, and clinical diagnoses and treatments. Most studies (42.2%, 38/90) used AI for image-based analysis, followed by clinical diagnoses and treatments (24 studies). Each of the remaining two subcategories included 14 studies. Machine learning and deep learning have been extensively applied in the field of otorhinolaryngology. However, the performance of AI models varies and research challenges remain.
ABSTRACT
Coronary heart disease (CHD) is one of the severe health issues and is one of the most common types of heart diseases. It is the most frequent cause of mortality across the globe due to the lack of a healthy lifestyle. Owing to the fact that a heart attack occurs without any apparent symptoms, an intelligent detection method is inescapable. In this article, a new CHD detection method based on a machine learning technique, e.g., classifier ensembles, is dealt with. A two-tier ensemble is built, where some ensemble classifiers are exploited as base classifiers of another ensemble. A stacked architecture is designed to blend the class label prediction of three ensemble learners, i.e., random forest, gradient boosting machine, and extreme gradient boosting. The detection model is evaluated on multiple heart disease datasets, i.e., Z-Alizadeh Sani, Statlog, Cleveland, and Hungarian, corroborating the generalisability of the proposed model. A particle swarm optimization-based feature selection is carried out to choose the most significant feature set for each dataset. Finally, a two-fold statistical test is adopted to justify the hypothesis, demonstrating that the performance differences of classifiers do not rely upon an assumption. Our proposed method outperforms any base classifiers in the ensemble with respect to 10-fold cross validation. Our detection model has performed better than current existing models based on traditional classifier ensembles and individual classifiers in terms of accuracy, F 1, and AUC. This study demonstrates that our proposed model adds a considerable contribution compared to the prior published studies in the current literature.
Subject(s)
Coronary Disease/diagnosis , Databases, Factual , Diagnosis, Computer-Assisted , Machine Learning , Models, Cardiovascular , HumansABSTRACT
Cloud computing supports many unprecedented cloud-based vehicular applications. To improve connectivity and bandwidth through programmable networking architectures, Software-Defined (SD) Vehicular Network (SDVN) is introduced. SDVN architecture enables vehicles to be equipped with SDN OpenFlow switch on which the routing rules are updated from a SDN OpenFlow controller. From SDVN, new vehicular architectures are introduced, for instance SD Vehicular Cloud (SDVC). In SDVC, vehicles are SDN devices that host virtualization technology for enabling deployment of cloud-based vehicular applications. In addition, the migration of Virtual Machines (VM) over SDVC challenges the performance of cloud-based vehicular applications due the highly mobility of vehicles. However, the current literature that discusses VM migration in SDVC is very limited. In this paper, we first analyze the evolution of computation and networking technologies of SDVC with a focus on its architecture within the cloud-based vehicular environment. Then, we discuss the potential cloud-based vehicular applications assisted by the SDVC along with its ability to manage several VM migration scenarios. Lastly, we provide a detailed comparison of existing frameworks in SDVC that integrate the VM migration approach and different emulators or simulators network used to evaluate VM frameworks' use cases.
