Fast diagnosis of bowel activities

Gastrointestinal health conditions, such as diarrhea, are one of the primary reasons of mortality in children as every year almost 1.6 to 2.5 million deaths happen around the world because of it. Diagnosis of gastrointestinal conditions is invasive and expensive to diagnose. We develop a new computerized diagnosis of gastrointestinal health conditions. The method solves the difficulties of recording bowel sounds by using only one bowel sound segment to detect types of bowel activities. The method utilizes Time Series Gaussian Hamming Distance (TSGHD) features to improve the robustness of the classifier in the presence jitters, noises, and location errors. We show that the new method allow recording of bowel sounds more practical and provides clinical explanations of bowel sounds and bowel activities. The proposed approach is evaluated on bowel sounds for classification tasks: normal and abnormal. It shows very promising results: 85.7% of accuracy, 88.5% of sensitivity and 81.3% of specificity.