Capsule Endoscopy Classification Using InceptionV3

The Capsule Endoscopy has also been acclaimed as a quickly emerging, promising diagnostic modality in cases of pathologies of the GI Tract, as this modality permits the non-invasive visualization of the small intestine. Nevertheless, as is evident, evaluating the large number of images generated during the diagnostic procedure has proven to be an arduous and laborious task. This research aims to propose an InceptionV3 Deep Learning Architecture-based Automatic Classifier for diagnosing endoscopic images from Capsule Endoscopy, which is a Convolutional Neural Network and is capable of achieving the best outcomes in the image classification task. The proposed scheme will be trained on a vast number of labeled endoscopic images of the GI Tract, including various pathologies such as ulcers, polyps, bleeding, and normal areas. Additionally, image preprocessing techniques used in this proposed scheme include resizing, normalization, and augmentation to improve efficiency. The proposed scheme, InceptionV3, has a hierarchical feature abstraction mechanism. Hence, this proposed scheme is capable of categorizing images into various groups. Additionally, this proposed scheme will be evaluated using multiple metrics, including accuracy, precision, recall, and F1 score, which have shown significant improvements compared to existing State-of-the-art Machine Learning algorithms.