Author ORCID Identifier
Date of Award
Spring 6-1-2026
Document Type
Thesis (Undergraduate)
Department
Computer Science
First Advisor
James Mahoney
Abstract
The integration of immersive technologies into clinical training has opened new possibilities for scalable, feedback-driven skill acquisition outside of supervised clinical settings. Breast ultrasound represents a particularly compelling application domain: with breast cancer affecting approximately 1 in 8 women over a lifetime, ultrasound plays a critical role in lesion characterization and early detection, yet it is a highly operator-dependent modality that demands the concurrent development of procedural probe technique and perceptual image reasoning. Novice learners frequently encounter this modality with limited structured practice and no real-time feedback on probe pressure, spatial coverage, or clinical reasoning.
This work presents an extended reality (XR) ultrasound training simulation for early novice trainees and junior radiology residents. The system blends a virtual anatomical phantom with a physical phantom, preserving tactile probe interaction in an immersive environment that screen-based training cannot replicate. A custom 3D-printed probe casing is fitted over the Meta Quest 3 controller, embedded with a pressure sensor, enabling contact force measurement alongside full controller motion tracking for scanning interaction in real-time. The system offers multiple training modes covering probe handling technique, spatial understanding and mental reconstruction of ultrasound cross-sections, and lesion localization, each supported by guided feedback and interactive visualizations. Throughout, skill acquisition is further reinforced through embedded educational gamification designed to motivate deliberate practice and self-assessment. A spherical coverage grid, scanned slice map, and live motion analytics further support self-directed practice, reducing educator dependence and enabling learning at the trainee's own pace. Clinical decision-making is integrated through ACR BI-RADS lesion classification, requiring trainees to evaluate shape, margins, posterior features, echogenicity, and malignancy risk stratification, bridging procedural scanning skill with interpretive clinical reasoning.
To evaluate skill transfer, a between-subjects RCT was conducted with sixteen participants, comparing XR simulation to traditional instruction with standardized assessments for probe handling accuracy and tumor localization. The XR group demonstrated significantly improved performance on outcomes dependent on volumetric and spatial reasoning, with coronal plane placement error reduced by approximately two-thirds, posterior feature identification accuracy increased by approximately 40 percentage points, and BI-RADS classification accuracy increased by approximately 51 percentage points, alongside lower perceived workload, higher usability, and greater self-efficacy. The system's architecture is designed to generalize beyond breast ultrasound, with broader implications for clinical training and remote educator-in-the-loop medical learning environments.
Recommended Citation
Dilibal, Cinay, "Real-Time Feedback-Driven Extended Reality Training Simulation for Enhancing Ultrasound Procedure Skill Acquisition" (2026). Computer Science Senior Theses. 69.
https://digitalcommons.dartmouth.edu/cs_senior_theses/69
Included in
Computer Sciences Commons, Medical Biotechnology Commons, Medical Education Commons, Telemedicine Commons
