Browsing by Author "Solomon, Justin B"

Artifacts are known to reduce the quality of CT images and can affect statistical analysis and quantitative utility of those images. Motion artifact is a leading type of CT artifact, due to either voluntary or involuntary (respiratory and cardiac movements) causes. Currently, such artifacts, if present, are not quantified and monitored, nor are their dependencies on CT acquisition settings known. As a first step to address this gap, the aim of this study was to develop a neural network to detect and quantify motion artifacts in CT images. Training data were drawn from three sources and the pixels containing motion were segmented (Seg3D, University of Utah) and the segmentation masks used as the ground truth labels. A convolutional neural network (u-net) was trained to identify pixels containing motion. The model performance was assessed by correlating the percentage of voxels labeled as having motion in each slice of the pre-allocated testing data for the ground-truth and predicted segmentation masks, yielding a correlation coefficient of r = 0.43, as well as constructing ROC curves. A series-wise ROC curve had AUC = 0.94, and a slice-wise ROC curve had AUC = 0.80. The correlation coefficient and AUCs are expected to improve as more training data is added. This network has potential to be a useful clinical tool, enabling quality tracking systems to detect and quantify the presence of artifacts in the context of CT quality control.

Computed tomography (CT) is one the most used imaging modalities due to its advancements in acquisition time and diagnostic capability. There are Concerns regarding the radiation exposure and its associated cancer risks. Therefore, great effort in research has been focused on minimizing patient dose by generating low-dose simulations. These simulations can be achieved by adding artificial noise into raw projection. However, raw CT data is not easily accessible. The purpose of this master’s project is to validate a simplified noise addition tool that requires only CT images to produce simulated low-dose images. The noise addition tool method aims to generate artificial noise that is similar to real CT noise in terms of magnitude, texture, spatial nonstationary characteristics. The tool was evaluated in terms of noise magnitude and texture through phantom and patient images studies. The tool was incorporated into routine clinical CT protocol review to demonstrate its usefulness. The first study of evaluating the tool was performed by imaging phantoms. A thorax anthropomorphic phantom and a multi-sized image quality phantom were scanned at different dose levels and reconstruction settings. Noise magnitude in the simulated low-dose images was compared to the actual images. Furthermore, the noise texture was also assessed through the noise power spectrum (NPS). The second study of evaluating the tool was performed using patient images. The images were obtained at high and low doses from various CT examinations and different reconstruction settings. The noise addition tool simulated low-dose images from the original high-dose images. The noise magnitude of the simulated low-dose images was compared to the actual low-dose images. The utility of the noise addition tool was applied to dose reduction of multiple myeloma skeletal CT protocol. Data from patients who underwent multiple myeloma scanning were retrospectively analyzed for radiation dose and image quality assessment. The tool was used to simulate the patient images at different low-dose levels. A musculoskeletal radiologist assessed the simulated images to determine the lowest possible dose. From the phantom images study, the noise magnitude in the simulated low-dose images compared to the actual images had a relative difference of xx. Also, NPS was visually comparable. From the patient images study, the noise magnitude comparison had a relative difference of 2.38%. In addition, the effective radiation dose of the multiple myeloma CT protocol was reduced by xx. In conclusion, the noise addition tool produced simulated CT images with realistic noise properties from standard dose CT images. The tool demonstrated its potential in estimating reduced dose protocols.