QUiLT (Quantitative Ultrasound in Longitudinal Tissue Tracking): Stitching 2D images into 3D Volumes for Organ Health Monitoring
Volumetric property maps of organs (e.g., stiffness) are clinically significant and valu- able in diagnosing the onset and monitoring the progression of a large multitude of diseases such as autosomal dominant polycystic kidney disease (ADPKD) and chronic liver disease (CLD). Unlike 2D property maps...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2023
|
| Online Access: | https://hdl.handle.net/1721.1/150136 |
| _version_ | 1826205221259313152 |
|---|---|
| author | Chen, Melinda |
| author2 | Anthony, Brian W. |
| author_facet | Anthony, Brian W. Chen, Melinda |
| author_sort | Chen, Melinda |
| collection | MIT |
| description | Volumetric property maps of organs (e.g., stiffness) are clinically significant and valu- able in diagnosing the onset and monitoring the progression of a large multitude of diseases such as autosomal dominant polycystic kidney disease (ADPKD) and chronic liver disease (CLD). Unlike 2D property maps, 3D property maps allow for precise, consistent, and accurate longitudinal comparison because they eliminate the variabil- ities associated with the underlying image acquisition protocol. 3D organ reconstruc- tions provide a holistic view of the organ and serve as a foundation for generating 3D property maps. Existing methods for reconstructing 3D images of organs include MRI and CT. Ultrasound emerges as a viable alternative due to its low cost, portability, and ability for repeated use.
This thesis presents a workflow for generating 3D images of the kidney and liver from 2D ultrasound images. We augment a 1D ultrasound probe with a set of sensors that allow for its localization in 3D space. The pose information and 2D images are combined to generate accurate organ volumes, specifically for the kidney and liver. In ex vivo studies, our method reconstructs renal volumes with an accuracy that is comparable to the current clinical gold-standard, CT. The liver, however, poses additional challenges due to its size and location underneath the ribcage – multiple partial scans are required to capture the entire organ. These scans need to be co- registered to generate a complete volume. Unfortunately, the liver parenchyma lacks features, such as edges and corners, that are used by conventional image registration techniques. To circumvent this, we propose a set of quantitative features based on the liver vasculature and develop an algorithm to determine their correspondences in 3D space. The features serve as internal landmarks that are used to perform deformable registration on the partial scans to create a whole-organ view. We validate the proposed approach with simulated and in vivo data; the method consistently performs as well or better than existing registration methods. Our method aims to provide an easy and cost-effective way for clinicians to monitor organ property changes over time, thereby paving the way for early diagnosis and prevention of diseases. |
| first_indexed | 2024-09-23T13:09:16Z |
| format | Thesis |
| id | mit-1721.1/150136 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T13:09:16Z |
| publishDate | 2023 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1501362023-04-01T03:29:45Z QUiLT (Quantitative Ultrasound in Longitudinal Tissue Tracking): Stitching 2D images into 3D Volumes for Organ Health Monitoring Chen, Melinda Anthony, Brian W. Harvard-MIT Program in Health Sciences and Technology Volumetric property maps of organs (e.g., stiffness) are clinically significant and valu- able in diagnosing the onset and monitoring the progression of a large multitude of diseases such as autosomal dominant polycystic kidney disease (ADPKD) and chronic liver disease (CLD). Unlike 2D property maps, 3D property maps allow for precise, consistent, and accurate longitudinal comparison because they eliminate the variabil- ities associated with the underlying image acquisition protocol. 3D organ reconstruc- tions provide a holistic view of the organ and serve as a foundation for generating 3D property maps. Existing methods for reconstructing 3D images of organs include MRI and CT. Ultrasound emerges as a viable alternative due to its low cost, portability, and ability for repeated use. This thesis presents a workflow for generating 3D images of the kidney and liver from 2D ultrasound images. We augment a 1D ultrasound probe with a set of sensors that allow for its localization in 3D space. The pose information and 2D images are combined to generate accurate organ volumes, specifically for the kidney and liver. In ex vivo studies, our method reconstructs renal volumes with an accuracy that is comparable to the current clinical gold-standard, CT. The liver, however, poses additional challenges due to its size and location underneath the ribcage – multiple partial scans are required to capture the entire organ. These scans need to be co- registered to generate a complete volume. Unfortunately, the liver parenchyma lacks features, such as edges and corners, that are used by conventional image registration techniques. To circumvent this, we propose a set of quantitative features based on the liver vasculature and develop an algorithm to determine their correspondences in 3D space. The features serve as internal landmarks that are used to perform deformable registration on the partial scans to create a whole-organ view. We validate the proposed approach with simulated and in vivo data; the method consistently performs as well or better than existing registration methods. Our method aims to provide an easy and cost-effective way for clinicians to monitor organ property changes over time, thereby paving the way for early diagnosis and prevention of diseases. Ph.D. 2023-03-31T14:34:56Z 2023-03-31T14:34:56Z 2023-02 2023-02-14T20:01:29.581Z Thesis https://hdl.handle.net/1721.1/150136 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Chen, Melinda QUiLT (Quantitative Ultrasound in Longitudinal Tissue Tracking): Stitching 2D images into 3D Volumes for Organ Health Monitoring |
| title | QUiLT (Quantitative Ultrasound in Longitudinal Tissue Tracking): Stitching 2D images into 3D Volumes for Organ Health Monitoring |
| title_full | QUiLT (Quantitative Ultrasound in Longitudinal Tissue Tracking): Stitching 2D images into 3D Volumes for Organ Health Monitoring |
| title_fullStr | QUiLT (Quantitative Ultrasound in Longitudinal Tissue Tracking): Stitching 2D images into 3D Volumes for Organ Health Monitoring |
| title_full_unstemmed | QUiLT (Quantitative Ultrasound in Longitudinal Tissue Tracking): Stitching 2D images into 3D Volumes for Organ Health Monitoring |
| title_short | QUiLT (Quantitative Ultrasound in Longitudinal Tissue Tracking): Stitching 2D images into 3D Volumes for Organ Health Monitoring |
| title_sort | quilt quantitative ultrasound in longitudinal tissue tracking stitching 2d images into 3d volumes for organ health monitoring |
| url | https://hdl.handle.net/1721.1/150136 |
| work_keys_str_mv | AT chenmelinda quiltquantitativeultrasoundinlongitudinaltissuetrackingstitching2dimagesinto3dvolumesfororganhealthmonitoring |