Lecture Notes on Artificial Intelligence in Spine Surgery

Introduction

• Presenter: Brian Krueger
• Topic: Artificial Intelligence (AI) in spine surgery
• Focus: AI's role in deformity correction surgery

Initial Thoughts on AI in Surgery

• Common misconception: AI as robots performing surgery
• Reality: AI enhances human intelligence, not replace surgeons
• AI is hundreds of years away from independently performing surgery
• Intimidation factor with AI for many, including those without a tech background

Notable Perspectives

• Elon Musk: Caution about AI; potential need for regulatory oversight
• Ginny Rometty (Former CEO of IBM): AI enhances human intelligence, augments capabilities

Overview of Presentation

• Challenges in adult spinal deformity surgery
• AI's potential in spinal deformity surgery
• Basic principles of machine learning
• Case examples using predictive models

Challenges in Spinal Deformity Surgery

• Numerous variables in planning: pathology, patient anatomy, biomechanics, bone density, patient goals, surgeon experience
• Difficulty in isolating variables for study
• AI can potentially study relationships among these variables

Potential of AI in Spine Surgery

• AI replicates human intelligence and learns from data sets
• Capable of analyzing massive datasets which humans cannot
• Reduces user bias and uncovers new relationships

Machine Learning in Spine Surgery

Three Paradigms:

1. Supervised Learning: Uses labeled data to train algorithms

   • Example: Identifying tree types
   • Spine application: Predict bone density, lumbar lordosis from X-rays

2. Unsupervised Learning: Finds patterns in unlabeled data

   • Example: Grouping data by similarities
   • Spine application: Identifying problematic variable combinations

3. Reinforcement Learning: Learning through trial and error

   • Improves over time with feedback
   • Long-term potential for better algorithms

UNiD Platform (Medtronic's AI)

• Supports surgeon workflow in pre-operative planning
• Provides personalized implant solutions
• Collects data to refine predictive models

Algorithms in UNiD Platform

• Predicts post-operative outcomes (e.g., thoracic kyphosis, pelvic tilt)
• Different algorithms for adult deformity, pediatrics, and degenerative cases

Case Examples

1. 50-year-old woman

   • History of L4-5 fusion
   • Plan: L2-4 ALIF, L5-S1 ALIF, T10 to pelvis
   • Outcome: Corrected sagittal and coronal planes

2. 56-year-old man

   • Previous failed surgery by non-deformity surgeon
   • Plan: S1-S2 ALIF, L4 PSO, T12 to pelvis
   • Outcome: Improved alignment, but less correction from PSO

3. 72-year-old woman

   • Severe lumbar kyphoscoliosis
   • Plan: Interbody fusions, significant correction
   • Outcome: Significant improvement, later minor issues

Conclusion

• AI in spine surgery is about enhancing surgeon capabilities
• AI is in its early stages; involvement now is beneficial for future developments
• Machine learning comprises collective experience and knowledge

Questions