Three-Dimensional Virtual Reconstructions of Shoulder Movements Using Computed Tomography Images: Model Development

Yu‐Hee Kim, In Park, Soo Buem Cho, Seoyon Yang, Il Kim, Kyong-Ha Lee, Kwangnam Choi, Seung‐Ho HanInteractive Journal of Medical Research 2023 · DOI

Abstract

The human shoulder is one of the most complex joints in the body, characterized by a wide range of motion driven by an intricate system of muscles and bones. This paper details the development of a real-time 3D "digital twin" of the shoulder that simulates these dynamics. By integrating Computed Tomography (CT) data with professional animation tools and game engines, the researchers created an interactive application capable of visualizing six primary shoulder motions with anatomically validated muscle deformations across nine key muscle groups.

Introduction

The concept of a "human digital twin"—a virtual mirror of a person’s biology—holds significant potential for personalized medicine, ranging from diagnosing joint disorders to surgical planning. However, creating an accurate digital twin is challenging because the human body is a dynamic system of soft tissues rather than a rigid skeleton. Most existing skeletal models lack muscle movement data, a significant gap since skeletal movement cannot be fully understood without accounting for the muscles that drive it. This study addresses this gap by developing a muscle-inclusive visualization tool for the shoulder, one of the most complex and mobile joints in the human body.

Methods

The research pipeline bridged the gap between medical imaging and interactive media through four key stages:

  1. Data Acquisition: Shoulder CT images in DICOM format were obtained from Ewha Womans University Seoul Hospital. The study also utilized the Digital Korean Project (2003–2007), a foundational database of Korean skeletal physical properties, as a benchmark for the digitalization process.
  2. 3D Modeling: DICOM files were converted to STL format and imported into Blender. Researchers performed manual segmentation to reconstruct three primary bones (scapula, clavicle, and humerus) and nine key muscles.
  3. Animation Logic: To ensure realistic movement, Blender’s Controller function was used for bone-muscle coordination, while the Shapekey function simulated the nuanced contraction, relaxation, and rotation of muscle tissue during joint movement.
  4. Interactive Application: The assets were exported to the Unity engine, where a user interface, camera controls, and interactive visibility toggles were implemented. "Collision boxes" were used to maintain software responsiveness during real-time interaction with complex polygonal units.
Anatomical reconstruction of the deltoid muscle on the skeletal frame.

Key Results

The resulting Unity-based application provides a "live" look at musculoskeletal mechanics that static scans cannot offer. Key achievements include:

  • Comprehensive Movement Library: The model successfully simulates six major shoulder movements: flexion, extension, abduction, adduction, internal rotation, and external rotation.
  • Anatomical Detail: The system includes high-fidelity reconstructions of nine specific muscles: the coracobrachialis, deltoid, infraspinatus, latissimus dorsi, pectoralis major, subscapularis, supraspinatus, teres major, and teres minor.
  • Real-Time Interaction: Users can rotate, zoom, and toggle the visibility of individual components in real-time, allowing for a deep-tissue view of musculoskeletal mechanics.
Visualization of the nine key shoulder muscles reconstructed in the 3D model.

Discussion and Insights

The core innovation of this model is the inclusion of agonistic muscles. Visualizing how muscles like the subscapularis or deltoid deform during rotation provides significantly more clinical value than a bone-only model. Furthermore, the animation is kinesiology-driven; alignment and rotation angles were programmed based on established textbooks to ensure the virtual movement reflects actual human range of motion. The "Peel-Back" UI allows users to see deep-tissue muscles normally obscured by larger groups, making it a powerful tool for education and clinical explanation.

Limitations and Future Work

While the model is a significant advancement, it currently relies on manual segmentation and pre-set animations. Future work will focus on:

  • Automation: Developing algorithms to automatically process patient-specific CT data, reducing the need for manual reconstruction.
  • Diagnostic Tools: Implementing automated systems within the app to identify "normal" vs. "abnormal" ranges of motion.
  • Clinical Application: Testing the tool's efficacy in rehabilitation settings and arthroplasty (joint replacement) planning.
  • Life-Cycle Tracking: The ultimate goal is an "evolving" digital twin that can reflect changes in a patient's anatomy over their entire lifespan.

Summary updated

BibTeX
@article{kim2023three,
  author = {Yu‐Hee Kim and In Park and Soo Buem Cho and Seoyon Yang and Il Kim and Kyong-Ha Lee and Kwangnam Choi and Seung‐Ho Han},
  title = {{Three-Dimensional Virtual Reconstructions of Shoulder Movements Using Computed Tomography Images: Model Development}},
  journal = {Interactive Journal of Medical Research},
  year = {2023},
  doi = {10.2196/48381},
}
Figure