Prediction of femoral head collapse in osteonecrosis using deep learning segmentation and radiomics texture analysis of MRI

Shihua Gao ¹, Haoran Zhu ², Moshan Wen ²

⁰Department of Orthopaedics, Zhongshan Hospital of Traditional Chinese Medicine Affiliated to Guangzhou University of Traditional Chinese Medicine, Zhongshan, Guangdong, China.

Bmc Medical Informatics and Decision Making +

|

November 1, 2024

View abstract on PubMed

Summary

This study developed an automated pipeline using MRI radiomics to predict femoral head collapse in osteonecrosis of the femoral head (ONFH). The developed model shows promising results for aiding clinical decisions in ONFH management.

Area Of Science

Radiomics and Medical Imaging
Artificial Intelligence in Healthcare
Orthopedic Surgery

Background

Femoral head collapse is a critical turning point in osteonecrosis of the femoral head (ONFH) progression.
Accurate prediction of this event is crucial for timely treatment and improved patient outcomes.

Purpose Of The Study

To develop an automated pipeline for predicting femoral head collapse in ONFH using magnetic resonance imaging (MRI) radiomics.
To assess the performance and clinical utility of a machine learning model for ONFH prognosis.

Main Methods

A deep learning model (nnU-Net) was trained for automatic lesion segmentation on T1-weighted MRI scans.
107 radiomics features were extracted, and feature selection was performed to identify key predictors.
A machine learning model (LightGBM) was trained for prognosis prediction using selected radiomics features.

Main Results

The segmentation model achieved a Dice similarity coefficient of 0.848 and a Hausdorff distance of 3.794.
The LightGBM prognosis prediction model demonstrated an Area Under the Curve (AUC) of 0.851, with 76.5% accuracy.
Decision curve analysis indicated favorable clinical utility for the developed model.

Conclusions

An automated pipeline for predicting femoral head collapse in ONFH using MRI radiomics has been successfully developed.
The radiomics-based approach shows acceptable performance and potential for assisting in ONFH treatment decision-making.
Further research is needed to validate its clinical applicability.

Keywords:

Deep learning MRI Machine learning Osteonecrosis of the femoral head Radiomics