Pretreatment Spatially Aware Magnetic Resonance Imaging Radiomics Can Predict Distant Brain Metastases (DBMs) After...

Area of Science:

• Radiology
• Oncology
• Medical Physics

Background:

• Stereotactic radiosurgery/radiation therapy (SRS/SRT) is a common treatment for brain metastases.
• Development of distant brain metastases (DBMs) in untreated brain areas remains a significant challenge.
• Predictive models are needed to identify patients at risk for DBMs.

Purpose of the Study:

• To develop a spatially aware radiomic signature for modeling time-to-DBM development.
• To leverage pretreatment magnetic resonance imaging (MRI) and radiation therapy dose distribution data.
• To compare radiomic signature performance against clinical factors.

Main Methods:

• Retrospective analysis of 105 patients treated with SRS/SRT for brain metastases.
• Extraction of 3D radiomic features from multiparametric MRI within isodose regions of interest (ROIs) and gross target volume (GTV).
• Cox proportional hazards (CPH) modeling and Kaplan-Meier analysis incorporating radiomic and clinical features.

Main Results:

• Radiomic models achieved a higher concordance index (c-index) (0.63) than clinical models (0.49).
• Combined radiomic and clinical models yielded the highest c-index (0.69).
• Radiomic features, particularly from peritumoral regions, significantly stratified patients by DBM development risk (P = .00007).

Conclusions:

• Radiomic features from pretreatment MRI and dose maps can effectively model time-to-DBM development after SRS/SRT.
• This approach surpasses traditional clinical factors in predictive accuracy.
• The findings suggest the peritumoral environment's role in DBM development and may aid in personalized treatment strategies.