Deep learning applications in orthopaedics: a systematic review and future directions

Area of Science:

• Orthopedic Surgery
• Artificial Intelligence
• Deep Learning

Background:

• Artificial intelligence (AI) and deep learning (DL) have rapidly gained interest in orthopedics.
• Previous research explored AI applications in areas like radiographic assessment and bone tumor diagnosis.

Purpose of the Study:

• To systematically review current literature on AI and DL tools in orthopedics.
• To identify prevalent AI tools used in risk assessment, outcome assessment, imaging, and basic science.

Main Methods:

• Systematic review of 595 studies from PubMed, EMBASE, and Google Scholar (Jan 2020 - Oct 2023).
• Included studies focused on radiographic assessment, spine surgery, outcome assessment, orthopedic education, and basic science.
• Meta-analysis used random effects to pool primary outcomes like diagnostic accuracy and reporting standards.

Main Results:

• Convolutional neural networks (73%) were the most frequently used machine learning architecture among 185 algorithms.
• AI tools primarily aimed to enhance diagnostic accuracy and speed, achieving these results in 62% of applications.
• 53 distinct imaging methods were employed for radiographic assessments.

Conclusions:

• High heterogeneity in methodology, terminology, and outcomes across studies was observed.
• Variations may lead to overestimation of diagnostic accuracy for DL algorithms in medical imaging.
• There is an urgent need for standardized AI-specific guidelines in orthopedic research.