Radiomics to predict high risk features in intraocular retinoblastoma.

Area of Science:

• Ophthalmology
• Radiology
• Oncology

Background:

• Detecting high-risk characteristics in intraocular retinoblastoma is crucial for assessing metastatic risk.
• Accurate risk stratification aids in treatment planning and prognosis for retinoblastoma patients.

Purpose of the Study:

• To evaluate the efficacy of preoperative Magnetic Resonance Imaging (MRI) radiomics in identifying high-risk histopathological features of retinoblastoma.
• To develop predictive models for optic nerve and choroidal invasion using MRI radiomics data.

Main Methods:

• A pilot study involving 31 eyes of 29 patients with retinoblastoma.
• Extraction and analysis of 2142 MRI radiomic features.
• Development of predictive models for optic nerve and choroidal invasion using selected radiomic features.

Main Results:

• The model for optic nerve invasion achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.85.
• The model for choroidal invasion achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.86.
• Specific radiomic features, including GLCM correlation and kurtosis of the optic nerve, and sphericity of the vitreous, were identified as significant predictors.

Conclusions:

• Preoperative MRI radiomics shows promise in predicting high-risk features in retinoblastoma.
• Radiomics may assist in guiding clinical decisions regarding globe salvage versus enucleation.
• Larger studies are needed to validate and refine these predictive models for clinical application.