University of Twente Student Theses

Login

Prediction of the hiatus esophagus location based on external landmarks and biometrics

Enthoven, J.J.M (2019) Prediction of the hiatus esophagus location based on external landmarks and biometrics.

Full text not available from this repository.

Full Text Status:Access to this publication is restricted
Abstract:Introduction – Malposition of a nasogastric tube (NGT) can have serious consequences. Unnoticed placements in the lungs brings the patient at risk and may lead to IC transfers with possible lethal endings. The current conformation methods, do not provide enough certainty of correct placement in clinical practice. The Stomach Tube Checker Project team is developing a device which will give information about the correct position of the NGT tip. One of the goals is the development of a sensor technique, which predicts a safe region to place the NGT in. This thesis has the focus on designing a prediction algorithm to define the location of the hiatus esophagus, the entry point of the stomach. It determines if the location of the of the hiatus esophagus can be predicted using external landmarks and biometrics in CT-scans. Requirements – Main requirement is to develop a prediction algorithm with a accuracy of 20 mm. Method – The data set consisting of external landmarks locations and biometrics were used as input parameters. The external landmarks were selected manually from CT-scans. The data set was randomly divided in a training group and a test group. The training group is again divided in a training set and a validation set. Three different algorithms were designed and compared to predict the location of the hiatus esophagus. The first algorithm uses the average locations of the training set as predicted location. The second algorithm describes a multiple linear regression and the third algorithm is an artificial neural network. Each algorithm calculates the deviation between the predicted location and the true location of the hiatus esophagus, which is described by the Euclidean distance. Results – The data set consists of 148 cases: 18 cases were randomly selected for the test group. 130 cases were selected for the training group, of which 30 cases are used for validation and 100 for the actual training. The multiple linear regression algorithm had the most promising results with a median Euclidean distance of 15 mm. The test group gave with the same algorithm a Euclidean distance of 18 mm, which meets the requirement of 20 mm for the prediction accuracy. Conclusion – An algorithm using a multiple linear regression gives the highest precision by predicting the location of the hiatus esophagus based on biometrics and external landmarks in CT-scans. This thesis confirms that it is possible to predict the location of the hiatus esophagus based on biometrics and external landmarks in CT-scans.
Item Type:Essay (Master)
Faculty:TNW: Science and Technology
Subject:44 medicine, 50 technical science in general
Programme:Technical Medicine MSc (60033)
Link to this item:http://purl.utwente.nl/essays/77819
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page