11/11/2024
By Debora Antonio
Date: Monday, Nov. 18, 2024
Time: 3 to 4:30 p.m.
Location: Virtual defense via Zoom. Those interested in attending should contact MS candidate Debora Antonio at least 24 hours prior to the defense to request access to the meeting.
Committee Members:
- Advisor: Konrad Nesteruk, Ph.D., Assistant Professor of Radiation Oncology, Massachusetts General Hospital
- Atchar Sudhyadhom, Ph.D., DARB, Assistant Professor of Radiation Oncology, Harvard Medical School
- Erno Sajo, Ph.D., Professor, Department of Physics and Applied Physics, UMass Lowell
- Gregory Sharp, Ph.D., DABR, Associate Professor of Radiation Oncology, Harvard Medical School
- Hengyong Yu, Ph.D., Professor, Department of Electrical & Computer Engineering, UMass Lowell
- Romy Guthier, Ph.D., Assistant Professor, Department of Physics, UMass Lowell
Abstract:
The pursuit for evermore higher medical image quality, with higher resolution and higher contrast, while minimizing radiation exposure to the patient, has become a highly sought-after topic. Therefore, it is crucial for research and development of new technological advances that tools capable of simulating images that closely resemble real human medical images are developed and provided.
The aim of this project was to develop an open-access toolkit for rapidly generating simultaneously realistic CT scans and low-field MR images of the abdominal region, based on real patient data, while employing an XCAT phantom geometry. The generated medical images are saved in DICOM format.
The developed code was written in MATLAB and requires a user-defined XCAT phantom geometry as input. The developed script generates simultaneously realistic 120keV CT scans and 0.35T MR images based on real patient data. The script then generates two new files, one with HU and another with MR intensities. For each voxel of the XCAT phantom, the corresponding organ is identified, and two new files are generated, containing the respective HU values or MR intensities. These HU values and MRI intensities were derived from actual patient CT scans and low-field MR images. The noise within both image modalities was assumed to be Gaussian, and therefore it follows a normal distribution, where the mean values and standard deviation values are based on patient images. Statistical data associated with the quantum noise for both image modalities is provided and compared to real patient data. Moreover, the newly generated images are provided in raw 32-bit format and can be saved in DICOM format along with a header containing the information on the XCAT slice number to which it corresponds.
The open-access toolkit allows the rapid production of realistic medical images for user-defined XCAT phantom geometries. Despite the existence of other simulation tools for CT scans and low-field MR image generation, this script has the advantage of not only generating both image modalities simultaneously in a short time, but also it does not require substantial computational power, nor does it require advanced programming skills.