Wireless Capsule Endoscopy Image Analysis using Deep Learning Techniques

Abstract

Worldwide, Wireless Capsule Endoscopy (WCE) is a useful method for doing easy newlinescreening and examining digestive system ailments within the intestines, several con- newlinesiderations, including efficiency, security, acceptance, and performance, make adap- newlinetion difficult and widespread. Furthermore, the computerized examination of the WCE newlineinformation is critical for detecting anomalies. Various vision-based and machine- newlinelearning technologies address such issues. However, they want more improvements newlineand do not provide an acceptable level of accuracy. To overcome this issue, first, we newlineintroduced a novel MobileNetV2-BiLSTM-based method for automated wireless capsule endoscopy image classification that distinguishes between affected and ordinary newlinepictures. BIR, which is a hybrid of MobileNetV2 and Bi-LSTM, uses MobileNetV2 for newlineclassification and Bi-LSTM for efficiency optimization. Regardless of data improve- newlinement, the implementation is handled. newlineAdditionally, WCE images typically contain a high number of tiny dots that are dif- newlineficult to identify. Therefore, we introduced a new Lenet-5-based deep learning method newlinefor gastrointestinal tract classification and identification of wireless capsule endoscopy. newlineAfter that, features are extracted using DeepLap v3+. Finally, the extracted features newlineare classified by utilizing Lenet-5. Extensive experimental testing on a standard dataset demonstrates the system s success, which is then compared to existing state-of-the-art methodologies. newlineFurthermore, deep learning will eventually be used to expand this research, to clas- newlinesify and identify the WCE. We developed an effective deep learning-based SE-ResNet newlinefor classifying gastrointestinal disorders. Initially, noises are eliminated in the im- newlineages using a median filter. After that images are extracted by using the DenseNet-121 newlinemethod. Applying the EWOA, pick the necessary characteristics following extracting features. Lastly, we suggest a SE-ResNet strategy using the Bald Eagle Searching newlineoptimization technique for classifying 8 categorie