東華大學圖書館 |

The Scattered Radiation Display System: A New Paradigm for Staff Member Dose Management and Training for Fluoroscopically-Guided Interventional Procedures Using Virtual Reality, Monte Carlo Simulations, and Artificial Intelligence.

Record Type:	Electronic resources : Monograph/item
Title/Author:	The Scattered Radiation Display System: A New Paradigm for Staff Member Dose Management and Training for Fluoroscopically-Guided Interventional Procedures Using Virtual Reality, Monte Carlo Simulations, and Artificial Intelligence./
Author:	Troville, Jonathan Louis.
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2022,
Description:	209 p.
Notes:	Source: Dissertations Abstracts International, Volume: 84-01, Section: B.
Contained By:	Dissertations Abstracts International84-01B.
Subject:	Medical imaging. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29210520
ISBN:	9798834010692

The Scattered Radiation Display System: A New Paradigm for Staff Member Dose Management and Training for Fluoroscopically-Guided Interventional Procedures Using Virtual Reality, Monte Carlo Simulations, and Artificial Intelligence.
Troville, Jonathan Louis.

The Scattered Radiation Display System: A New Paradigm for Staff Member Dose Management and Training for Fluoroscopically-Guided Interventional Procedures Using Virtual Reality, Monte Carlo Simulations, and Artificial Intelligence. - Ann Arbor : ProQuest Dissertations & Theses, 2022 - 209 p.

Source: Dissertations Abstracts International, Volume: 84-01, Section: B.

Thesis (Ph.D.)--State University of New York at Buffalo, 2022.

This item must not be sold to any third party vendors.

Fluoroscopically-guided interventional (FGI) procedures used for treating vascular pathologies, such as aneurysms and stenoses, can be long in duration leading to an increased risk of deterministic and stochastic effects. To mitigate the onset of deterministic skin effects post-procedure, our group developed the real-time patient skin Dose Tracking System (DTS), which is now a product licensed to Canon available for distribution with a variety of biplane and single-plane angiography systems. In addition to an increased risk to the patient, prolonged procedures increase risk to staff members. While interventionalists are typically stationary, and remain behind tableside shielding, during the course of FGIs, ancillary staff members, such as the technologist, nurse, and anesthesiologist, may move around and neglect additional shielding available in the room.The as low as reasonably achievable (ALARA) principle maintains that absorbed radiation dose for both the patient and staff members should be minimized while concurrently ensuring proper treatment outcome. Standard methods exist for monitoring staff dose levels either during or after procedures in the form of real-time ionization chamber dosimeters or thermoluminescent dosimeters (TLDs). However, these dosimeters are typically worn at the collar level and waist level, only providing point-dose estimations. With the complex dynamics of scatter distributions in the interventional room, in addition to the inability to view x-ray radiation with the naked eye, the staff remain unaware of how to optimize their location along with geometric and exposure parameters. xvThe primary work presented in this dissertation is a comprehensive software system capable of monitoring changes to scatter distributions during FGIs, as well as extra-procedure training, called the scattered radiation display system (SDS). The SDS dose management module (DM) functions in real-time via interfacing to an angiography system with a controller area network (CAN) bus. Hexadecimal data packets containing geometric and exposure information are fed off of the CAN to the SDS directly and utilized to properly select a color-coded scatter distribution for display. The displayed distribution is derived from pre-calculated matrices using EGSnrc Monte Carlo (MC). Two methods are compared and contrasted in this dissertation for providing a sufficient amount of information to determine the displayed distribution: 1) a discrete look-up-table consisting of 3D MC pre-calculated distributions which correspond to configurations of exposure and geometric parameters known to alter the spatial shape of the distribution; 2) a regressional deep neural network (DNN) which predicts the displayed distribution using a dataset of MC pre-calculated distributions. In addition to providing the in-room scatter distribution, the SDS consists of a 3D virtual background rendered using OpenGL. Stereolithographic (STL) meshes were created in Make Human for the patient models, whereas a Microsoft Kinect V2 depth sensing camera and the Kinect Fusion Algorithm were used for generating an STL mesh of a C-Arm gantry in our lab. In addition, the table model is an STL mesh representing the catheterization table in the clinical environment. For use in the clinic, a ceiling-to-floor perspective is maintained in the SDS where the displayed distribution is a 2D horizontal slice corresponding to waist-level height.Two methods are discussed for providing positional references in the SDS for staff members. The first is a simple display which consists of the scatter distribution, virtual background, and distance reference indicators. The second is a display which consists of the scatter distribution, virtual background, and staff member position indicators. The latter method uses depth information from a Microsoft Kinect V2 depth camera interfaced to a robot operating system (ROS) wrapper for body tracking to update position indicators for each staff member involved with an FGI procedure. Calibration and accuracy of the Kinect V2 are discussed using a pinhole camera model with a lens distortion correction in OpenCV. Additionally, a linear (rigid-body) transformation algorithm was developed for translating the positional information derived using the ROS wrapper to on-screen coordinates in the SDS.Using body tracking with staff member position indicators in the SDS opens up an avenue for providing individualized dose-rate estimates during procedures in addition to cumulative dose estimates post-procedure for various locations on each staff member's body. In this dissertation, a convolutional neural network (CNN) is presented for human recognition with the intent of associating specific IDs, which correspond to the unique staff members in the room, to the staff member position indicators in the SDS. The CNN utilizes binary body silhouette masks of individuals for training, which are derived from raw depth frames using connected component analysis. SDS dose rate reports were developed which present each staff member with cumulative dose estimates at the torso level, collar level, and eye level after a procedure is complete.To improve the accuracy of doses displayed in the SDS, umbra regions produced via attenuation by important objects in the room (i.e. staff members, C-Arm gantry, and ceiling-mounted shield) are modeled. In order to track translational changes for the ceiling-mounted shield position, a features from accelerated segment test (FAST) based machine learning algorithm was implemented in OpenCV with an ROS wrapper for identifying and tracking the flange connecting the shield to the mounting arm. We also discuss in detail a method for automatically measuring the ceiling-mounted shield's orientation through co-registration of the RGB and depth feeds from the Kinect V2. Similar to the C-Arm gantry, the Microsoft Kinect V2 and Kinect Fusion Algorithm were used for generating an STL mesh of the ceiling-mounted shield, which is included in the SDS' virtual background. Algorithms for modeling the umbra regions are discussed and evaluated for accuracy. Full SDS validation is discussed, where a solid water phantom is used to validate scatter distributions and a beam spectrum.In addition to the SDS DM, this dissertation discusses the development of a training module (TM), which can provide the ability to perform simulated procedures as well as post-procedure review. This system makes use of virtual reality (VR) via an Oculus Go headset and controller interfaced to a workstation on which the SDS is installed. Unity3D is implemented for rendering the 3D environment, which is presented to the viewer wearing the Oculus Go headset. Three options are discussed for providing/enhancing training: 1) intra-procedural viewing; 2) post-procedural analysis; 3) parameter modification with a graphical user interface (GUI). The results presented for real-time performance and accuracy indicate that the SDS DM is feasible for providing a rapid and intuitive visual feedback for staff members during procedures. The SDS, in addition to a VR training system, would provide enhanced radiation awareness and improved staff dose levels during FGI procedures.

ISBN: 9798834010692Subjects--Topical Terms:

3172799
Medical imaging.
Subjects--Index Terms:

Deep learning

The Scattered Radiation Display System: A New Paradigm for Staff Member Dose Management and Training for Fluoroscopically-Guided Interventional Procedures Using Virtual Reality, Monte Carlo Simulations, and Artificial Intelligence.
LDR:08703nmm a2200385 4500 001 2350422
005 20221020125800.5
008 241004s2022 ||||||||||||||||| ||eng d
020 $a 9798834010692
035 $a (MiAaPQ)AAI29210520
035 $a AAI29210520
040 $a MiAaPQ $c MiAaPQ
100 1 $a Troville, Jonathan Louis. $0 (orcid)0000-0002-5090-5698 $3 3689900
245 1 4 $a The Scattered Radiation Display System: A New Paradigm for Staff Member Dose Management and Training for Fluoroscopically-Guided Interventional Procedures Using Virtual Reality, Monte Carlo Simulations, and Artificial Intelligence.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2022
300 $a 209 p.
500 $a Source: Dissertations Abstracts International, Volume: 84-01, Section: B.
500 $a Includes supplementary digital materials.
500 $a Advisor: Bednarek, Daniel R.
502 $a Thesis (Ph.D.)--State University of New York at Buffalo, 2022.
506 $a This item must not be sold to any third party vendors.
520 $a Fluoroscopically-guided interventional (FGI) procedures used for treating vascular pathologies, such as aneurysms and stenoses, can be long in duration leading to an increased risk of deterministic and stochastic effects. To mitigate the onset of deterministic skin effects post-procedure, our group developed the real-time patient skin Dose Tracking System (DTS), which is now a product licensed to Canon available for distribution with a variety of biplane and single-plane angiography systems. In addition to an increased risk to the patient, prolonged procedures increase risk to staff members. While interventionalists are typically stationary, and remain behind tableside shielding, during the course of FGIs, ancillary staff members, such as the technologist, nurse, and anesthesiologist, may move around and neglect additional shielding available in the room.The as low as reasonably achievable (ALARA) principle maintains that absorbed radiation dose for both the patient and staff members should be minimized while concurrently ensuring proper treatment outcome. Standard methods exist for monitoring staff dose levels either during or after procedures in the form of real-time ionization chamber dosimeters or thermoluminescent dosimeters (TLDs). However, these dosimeters are typically worn at the collar level and waist level, only providing point-dose estimations. With the complex dynamics of scatter distributions in the interventional room, in addition to the inability to view x-ray radiation with the naked eye, the staff remain unaware of how to optimize their location along with geometric and exposure parameters. xvThe primary work presented in this dissertation is a comprehensive software system capable of monitoring changes to scatter distributions during FGIs, as well as extra-procedure training, called the scattered radiation display system (SDS). The SDS dose management module (DM) functions in real-time via interfacing to an angiography system with a controller area network (CAN) bus. Hexadecimal data packets containing geometric and exposure information are fed off of the CAN to the SDS directly and utilized to properly select a color-coded scatter distribution for display. The displayed distribution is derived from pre-calculated matrices using EGSnrc Monte Carlo (MC). Two methods are compared and contrasted in this dissertation for providing a sufficient amount of information to determine the displayed distribution: 1) a discrete look-up-table consisting of 3D MC pre-calculated distributions which correspond to configurations of exposure and geometric parameters known to alter the spatial shape of the distribution; 2) a regressional deep neural network (DNN) which predicts the displayed distribution using a dataset of MC pre-calculated distributions. In addition to providing the in-room scatter distribution, the SDS consists of a 3D virtual background rendered using OpenGL. Stereolithographic (STL) meshes were created in Make Human for the patient models, whereas a Microsoft Kinect V2 depth sensing camera and the Kinect Fusion Algorithm were used for generating an STL mesh of a C-Arm gantry in our lab. In addition, the table model is an STL mesh representing the catheterization table in the clinical environment. For use in the clinic, a ceiling-to-floor perspective is maintained in the SDS where the displayed distribution is a 2D horizontal slice corresponding to waist-level height.Two methods are discussed for providing positional references in the SDS for staff members. The first is a simple display which consists of the scatter distribution, virtual background, and distance reference indicators. The second is a display which consists of the scatter distribution, virtual background, and staff member position indicators. The latter method uses depth information from a Microsoft Kinect V2 depth camera interfaced to a robot operating system (ROS) wrapper for body tracking to update position indicators for each staff member involved with an FGI procedure. Calibration and accuracy of the Kinect V2 are discussed using a pinhole camera model with a lens distortion correction in OpenCV. Additionally, a linear (rigid-body) transformation algorithm was developed for translating the positional information derived using the ROS wrapper to on-screen coordinates in the SDS.Using body tracking with staff member position indicators in the SDS opens up an avenue for providing individualized dose-rate estimates during procedures in addition to cumulative dose estimates post-procedure for various locations on each staff member's body. In this dissertation, a convolutional neural network (CNN) is presented for human recognition with the intent of associating specific IDs, which correspond to the unique staff members in the room, to the staff member position indicators in the SDS. The CNN utilizes binary body silhouette masks of individuals for training, which are derived from raw depth frames using connected component analysis. SDS dose rate reports were developed which present each staff member with cumulative dose estimates at the torso level, collar level, and eye level after a procedure is complete.To improve the accuracy of doses displayed in the SDS, umbra regions produced via attenuation by important objects in the room (i.e. staff members, C-Arm gantry, and ceiling-mounted shield) are modeled. In order to track translational changes for the ceiling-mounted shield position, a features from accelerated segment test (FAST) based machine learning algorithm was implemented in OpenCV with an ROS wrapper for identifying and tracking the flange connecting the shield to the mounting arm. We also discuss in detail a method for automatically measuring the ceiling-mounted shield's orientation through co-registration of the RGB and depth feeds from the Kinect V2. Similar to the C-Arm gantry, the Microsoft Kinect V2 and Kinect Fusion Algorithm were used for generating an STL mesh of the ceiling-mounted shield, which is included in the SDS' virtual background. Algorithms for modeling the umbra regions are discussed and evaluated for accuracy. Full SDS validation is discussed, where a solid water phantom is used to validate scatter distributions and a beam spectrum.In addition to the SDS DM, this dissertation discusses the development of a training module (TM), which can provide the ability to perform simulated procedures as well as post-procedure review. This system makes use of virtual reality (VR) via an Oculus Go headset and controller interfaced to a workstation on which the SDS is installed. Unity3D is implemented for rendering the 3D environment, which is presented to the viewer wearing the Oculus Go headset. Three options are discussed for providing/enhancing training: 1) intra-procedural viewing; 2) post-procedural analysis; 3) parameter modification with a graphical user interface (GUI). The results presented for real-time performance and accuracy indicate that the SDS DM is feasible for providing a rapid and intuitive visual feedback for staff members during procedures. The SDS, in addition to a VR training system, would provide enhanced radiation awareness and improved staff dose levels during FGI procedures.
590 $a School code: 0656.
650 4 $a Medical imaging. $3 3172799
650 4 $a Artificial intelligence. $3 516317
653 $a Deep learning
653 $a Dose reduction
653 $a Fluoroscopically-guided intervention
653 $a Human recognition
653 $a Monte Carlo simulations
653 $a Software development
690 $a 0574
690 $a 0800
710 2 $a State University of New York at Buffalo. $b Medical Physics. $3 3689901
773 0 $t Dissertations Abstracts International $g 84-01B.
790 $a 0656
791 $a Ph.D.
792 $a 2022
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29210520