International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024
p-ISSN: 2395-0072
www.irjet.net
HYBRID CNN-LSTM MODEL FOR THE CLASSIFICATION OF WIRELESS CAPSULE ENDOSCOPY IMAGES FOR BLEEDING OR NORMAL DIAGNOSIS DIVYA BHARATHI.P1, RAMACHANDRAN.M2, RAJKUMAR.M3, RAJKUMAR.R. K4 1Assistant Professor Department of Computer Science and Engineering,
K.L.N. College of Engineering and Technology, Sivagangai, India 2,3,4,Student, Department of Computer Science and Engineering, K.L.N. College of Engineering and Technology, Sivagangai, India ----------------------------------------------------------------------***----------------------------------------------------------------------
Abstract-Wireless Capsule Endoscopy (WCE) has emerged
Key Words- Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Wireless Capsule Endoscopy (WCE), Deep Learning, detecting bleeding etc...
as a pivotal tool for diagnosing gastrointestinal disorders due to its non-invasive nature and ability to capture highresolution images throughout the digestive tract. However, the sheer volume of data generated by WCE procedures poses significant challenges for efficient analysis and interpretation. Automated classification of WCE images into clinically relevant categories, such as identifying the presence of bleeding, is essential for assisting medical professionals in timely diagnosis and intervention. In this study, we propose a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model for the classification of WCE images into two classes: bleeding and normal. The CNN component serves as a feature extractor, leveraging its ability to capture spatial dependencies within images, while the LSTM component handles the temporal dynamics inherent in sequences of images captured during the endoscopic procedure. Our hybrid model is trained on a large dataset of annotated WCE images, utilizing transfer learning techniques to leverage pre-trained CNN architectures for feature extraction. Subsequently, the features extracted by the CNN are fed into the LSTM network, which learns the temporal dependencies between consecutive frames of WCE images. To evaluate the performance of our proposed model, extensive experiments are conducted on a diverse dataset comprising WCE images from various patients with gastrointestinal conditions. The results demonstrate that our hybrid CNN-LSTM model achieves superior classification accuracy compared to standalone CNN or LSTM models. Furthermore, our model exhibits robustness to variations in image quality and illumination conditions commonly encountered in clinical settings. The proposed hybrid CNN-LSTM model holds great promise for enhancing the efficiency and accuracy of diagnosing gastrointestinal disorders through WCE. By automating the classification of WCE images into clinically relevant categories, such as detecting bleeding, our model can assist medical practitioners in making timely and informed decisions, ultimately improving patient outcomes and healthcare delivery in gastroenterology.
© 2024, IRJET
|
Impact Factor value: 8.226
1.INTRODUCTION Capsule endoscopy is a medical procedure used to record internal images of the gastrointestinal tract for use in disease diagnosis. Newer developments are also able to take biopsies and release medication at specific locations of the entire gastrointestinal tract. Unlike the more widely used endoscope, capsule endoscopy provides the ability to see the middle portion of the small intestine. It can be applied to the detection of various gastrointestinal cancers, digestive diseases, ulcers, unexplained bleedings, and general abdominal pains. After a patient swallows the capsule, it passes along the gastrointestinal tract, taking a number of images per second which are transmitted wirelessly to an array of receivers connected to a portable recording device carried by the patient. General advantages of capsule endoscopy over standard endoscopy include the minimally invasive procedure setup, ability to visualize more of the gastrointestinal tract, and lower cost of the procedure. In this context, the development of hybrid models combining convolutional neural networks (CNNs) and recurrent neural networks (RNNs), such as long shortterm memory (LSTM) networks, has garnered significant attention. These models leverage the spatial and temporal information present in WCE images, capturing both the intricate details of mucosal patterns and the sequential dynamics of peristalsis and lesion evolution. The process begins with the patient swallowing the capsule, which contains a tiny camera capable of capturing images. As the capsule moves through the GI tract, it continuously captures images of the intestinal lining. These images are wirelessly transmitted to an external recording device worn by the patient, where they are stored for subsequent analysis. Before analysis, WCE images undergo pre-
|
ISO 9001:2008 Certified Journal
|
Page 667