Integrated Path Planning System for Freehand Ultrasound Scanning
Name
lin-qalin-engineer-meche-2026-thesis.pdf
Size
10.06 MB
Format
Adobe PDF
Checksum (MD5)
40ec632fea1b23324eca93dc015255bf
Author(s)
Lin, Qian
Advisor(s)
Anthony, Brian W.
Date Issued
February 2026
Publisher
Massachusetts Institute of Technology
Abstract
Volumetric medical imaging offers richer anatomical representation than traditional two-dimensional modalities; however, reliable three-dimensional reconstruction from freehand ultrasound remains difficult due to the absence of inherent probe localization and the high dependence on operator expertise. While freehand ultrasound is widely adopted because of its cost-effectiveness, noninvasiveness, portability, and real-time imaging capability, these limitations hinder consistent volume completeness and increase inter-operator variability, particularly for non-expert users. This thesis presents a novel real-time acquisition-guidance and path-planning framework designed to assist users during freehand ultrasound scanning. The proposed framework explicitly optimizes incremental voxel coverage of the reconstructed volume, incorporating an occlusion-aware coverage objective to reduce under-sampled regions. Guidance is generated using a constrained set of probe motion primitives consisting of translation and rotation, enabling computationally efficient planning suitable for real-time interaction. Scan completion is determined automatically using a coverage saturation criterion (e.g., 95% target voxel coverage). The framework deliberately does not address probe contact force, tissue deformation modeling, speed-of-sound correction, or diagnostic plane classification, focusing instead on robust spatial coverage and real-time usability. The system integrates the path-planning module with a real-time visualization interface that provides instantaneous probe-movement cues and live image feedback, allowing continuous adaptation of user motion and rapid three-dimensional reconstruction during acquisition. This closed-loop guidance paradigm aims to improve image integrity, reduce user dependency, and support more reliable volumetric ultrasound acquisition. The methodology encompasses data acquisition, fast volumetric reconstruction, real-time visualization and guidance, and experimental validation. Evaluation is conducted using ultrasound phantoms and CT-label-based simulation, including pose-uncertainty stress testing. Quantitative performance is assessed using metrics such as voxel coverage recall, Dice similarity, coverage efficiency, and overconfidence under localization noise. This work contributes a prototype framework toward clinically viable, guided 3D freehand ultrasound imaging.
MIT Department
Massachusetts Institute of Technology. Department of Mechanical Engineering
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
