Diagnostic Accuracy of Artificial Intelligence-Based Algorithms in Automated Detection of Neck of Femur Fracture on a Plain Radiograph: A Systematic Review and Meta-analysis

Manish Raj ¹, Arshad Ayub ², Arup Kumar Pal ³

⁰Department of Orthopaedic, All India Institute of Medical Sciences, Deoghar, Jharkhand India.

Indian Journal of Orthopaedics +

|

May 2, 2024

View abstract on PubMed

Summary

Artificial intelligence (AI) algorithms show promise in diagnosing neck of femur (NOF) fractures on X-rays, with pooled sensitivity of 85% and specificity of 87%. These AI tools can assist clinicians, improving diagnostic efficiency and reducing workload.

Area Of Science

Radiology
Medical Imaging
Artificial Intelligence

Background

Neck of femur (NOF) fractures are a common and serious injury, particularly in elderly populations.
Accurate and timely diagnosis of NOF fractures is crucial for effective patient management and outcomes.
Plain radiography remains a primary imaging modality for NOF fracture detection, but interpretation can be challenging.

Purpose Of The Study

To systematically review and meta-analyze the diagnostic accuracy of artificial intelligence (AI)-based algorithms for identifying NOF fractures on plain radiographs.
To quantify the performance of AI algorithms using metrics such as sensitivity, specificity, and Youden index.

Main Methods

A comprehensive search of multiple databases (PubMed, Web of Science, Scopus, IEEE, Science Direct) was conducted up to July 2023.
Eligible studies included descriptive, analytical, or trial studies in English that evaluated AI algorithms for NOF fracture detection on X-rays.
Data extraction and risk of bias assessment were performed by two independent reviewer teams using CLAIM and QUADAS-2 criteria.

Main Results

Five studies met the inclusion criteria from 437 retrieved articles.
The pooled sensitivity for AI in diagnosing NOF fractures was 85%, with a pooled specificity of 87%.
The pooled Youden index was 0.73, with a positive likelihood ratio of 19.88 and a negative likelihood ratio of 0.17. AI showed comparable performance to human diagnosis with marginal heterogeneity (I² = 51%).

Conclusions

AI-based algorithms demonstrate significant potential as a diagnostic adjunct for NOF fracture detection on plain radiographs.
These AI tools can enhance diagnostic efficiency, saving clinicians time and effort.
Further integration of AI in radiology workflows could improve the accuracy and speed of NOF fracture diagnosis.

Keywords:

Artificial intelligence Fracture detection Neck of femur Plain radiograph