go-v.orgdoi

Development and Evaluation of a Machine Learning Approach for Application Validation of a Novel VR/AR Surgical Training Device of a Spinal Operation with Focus on Physics-Based Force Feedback

Sami Alkadri · eScholarship@McGill (McGill) · 2024

  • DOI 10.82308/9331
  • Publisher McGill University
  • Type dissertation
  • Issued 2024

Not cached yet. This record was fetched live for this request and is stored after the lookup. Reopening the page should serve the cached copy.

eScholarship@McGill (McGill) · McGill University · dissertation · 2024

Abstract

This thesis validates a physics-based virtual (VR) and augmented reality (AR) simulator for training in Oblique Lateral Lumbar Interbody Fusion (OLLIF) surgery, addressing the increasing demand for effective spinal surgical training. With over 80% of people experiencing back pain, leading to surgery, the precision and safety in spinal operations are paramount. Minimally Invasive Spinal Surgeries (MISS), although reducing recovery times, pose significant training challenges due to their complexity and reliance on advanced surgical skills.

Traditional training methods have proven inadequate, highlighted by the limitations of the "see one, do one, teach one" model and restricted resident working hours, necessitating innovative training solutions such as VR/AR surgical simulations. However, despite the encouraging initial outcomes of VR/AR surgical training systems, thorough validation studies are essential to advance their integration into surgical education programs. The project's main objective was achieved through a sequential validation approach, encompassing subjective assessments, machine learning analyses, and the examination of haptic feedback's role in surgical performance.

The first objective established the foundational validation of the simulator – defined by face, content, and construct validity – using both subjective and objective methods. Starting with subjectively assessing the resemblance of the simulated scenario to reality, a side-by-side comparison with a cadaveric surgery is conducted to further support face and content validity of the developed simulator. Face-validity is the extent to which the developed simulation mimics reality, whereas content-validity is the extent to which it is representative of the skills required to successfully complete the real surgery.

Afterwards, construct validity was established by extracting psychomotor data to derive surgical metrics that distinguish between skill levels. The second objective explored the application of artificial neural networks (ANNs) for surgical performance classification, developing a novel methodology for feature importance in multilayered ANNs and addressing challenges related to limited datasets through data augmentation and transfer learning. This objective also demonstrated the utility of combining neural network depth with traditional statistical analysis breadth for a holistic understanding of surgical expertise.

The third objective focused on the significance of accurate physics-based force feedback, particularly in the “gaining access” step of the OLLIF surgery, where visual feedback is limited. It introduced a novel approach to objectively measure haptic fidelity, highlighting how realistic forces directly influence surgical simulation performance and, by extension, training outcomes.The dissertation presents original contributions to surgical simulation validation, machine learning in surgical performance analysis, and the biomechanics of surgery, offering a comprehensive framework for future surgical simulator development and validation.

The methodologies developed have broader implications, paving the way for more effective training tools in high-risk fields. This work advocates for standardized guidelines in surgical simulation validation and emphasizes the necessity of incorporating accurate haptic feedback in training to enhance the safety and effectiveness of MISS

Open the resolved locator · 10.82308/9331

Metadata licence: OpenAlex metadata (CC0)

Registry
OpenAlex DOI index + API
Fetched
2026-11-20 03:24 UTC
Cache state
not cached yet (bulk-index+openalex)
Locator
http://escholarship.mcgill.ca/downloads/bc386r33n
Metadata licence
OpenAlex metadata (CC0)
Scope
Metadata as returned by the resolver; abstracts are never generated here.