CLASSIFICATION OF SMALL BOWEL OBSTRUCTION USING ABDOMINAL X-RAY IMAGES WITH A HYBRID APPROACH

Abstract

Small bowel obstruction (SBO) is a common and potentially fatal condition that occurs with the partial or complete blockage of the small bowel, hindering the transport of digestive content. The most common causes are adhesions following surgery, hernias, tumors, Crohn's disease, and intussusception. The patient experiences clinical presentation of pain in the abdomen, vomiting, distension, and constipation. Possessing an accurate and timely diagnosis is critical in preventing complications such as ischemia or perforation; appropriate investigations including abdominal X rays and CT scans are therefore necessary. With the advent of deep learning techniques in medical image analysis, there has been increasing encouragement in improving diagnostic accuracy. Earlier approaches were mostly focused on either single-feature extraction methods or individual classifiers. We propose a hybrid deep learning approach combining CNN and Histogram of Oriented Gradients (HOG) for feature extraction, followed by Support Vector Machines (SVMs) for classification. This hybrid approach combines both worlds in traditional machine learning and deep learning techniques. The accuracy achieved by the proposed model in classifying obstructed from non-obstructed cases is 97%. Our model was also compared against some renowned deep learning models, including InceptionV3, VGG16, and AlexNet, for the sake of performance evaluation. Experimental results showed that the hybrid model surpassed these state-of-the-art models in classification accuracy, thus indicating its potential application in clinical diagnostic workflows. In essence, this study has built a robust and reliable automated diagnostic system for detecting and classifying SBO from abdominal X-ray images. The resultant findings are very much contributory to AI-assisted diagnostic imaging and greatly support reducing the burden of work on health professionals while ensuring timely and accurate medical interventions. Keywords: Small Bowel Obstruction, Deep Learning, Convolutional Neural Networks (CNN), Support Vector Machine (SVM), Abdominal X-ray Images, Automated Diagnosis, Classificatio