Measuring and Predicting Image Noise of Computed Tomographic Examinations using statistical and deep learning approaches
Loading...
Links to Files
Permanent Link
Author/Creator
Author/Creator ORCID
Date
2020-01-20
Type of Work
Department
Computer Science and Electrical Engineering
Program
Computer Science
Citation of Original Publication
Rights
Distribution Rights granted to UMBC by the author.
This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see http://aok.lib.umbc.edu/specoll/repro.php or contact Special Collections at speccoll(at)umbc.edu
This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see http://aok.lib.umbc.edu/specoll/repro.php or contact Special Collections at speccoll(at)umbc.edu
Abstract
Computed Tomography (CT) plays an important role in diagnosing various diseases. However, CT image acquisition causes the risk of radiation exposure to patients. Thus, it is important to optimize imaging procedures so that the scan can be performed at the lowest dose. Lowering radiation dose impacts the image quality for visualization diagnosis. In practice, the assessment of CT image quality is still challenging due to its various imaging acquisition procedures, patient characteristics, image reconstruction kernel. In this theses we propose a Deep Learning approach using Convolution Neural Network performs the classification task for Image Quality Assessment (whether or not images are good for clinical diagnostics). The deep learning model uses residual network which is a type of convolutional neural network having skip layers and add layers. After training with quality rating labels from radiologists, the model predicts whether the Computed Tomography scan is good for diagnosis or the patient has to retake it with high accuracy (95% using synthetic images, 80% on clinal images). In this theses we also explained a statistical approach that uses an algorithm to find the region of interest and finally calculates the noise in the specific region of CT scans. Finally we analyze with statistical algorithms and models. The public Kaggle Data Science Bowl 2017, low dose lung cancer dataset, and UCSF 740 clinical dataset is used for running various experiments for noise classification on patient studies. Accuracy of the model and a statistical algorithm is used to evaluate the reliability of the results produced by the model.