Development of a Predictive Model for Heart Dose in Left Breast Radiotherapy Using Geometric Analysis

Abstract

This paper presents the development of a predictive model to correlate the mean heart dose (MHD) in left breast radiotherapy with treatment geometry, specifically aiming to minimize cardiac complications due to heart proximity. Given the importance of constraining MHD, we propose a methodology to quantify geometric arrangements using the Expansion Intersection Histogram (EIH) method. This approach involves progressive isotropic expansions of the target volume and mapping the overlap with the heart. From the resulting EIH graph, two key parameters—separation (S) and wrapping (W)—are derived, alongside the omolateral breast volume (OBV), which serve as inputs for a multivariate linear regression model. The study includes data from 19 breast cancer patients who received a treatment course of 15 fractions, with a breast dose of 40.5 Gy and a 48 Gy simultaneous integrated boost (SIB) planned via volumetric modulated arc therapy (VMAT). Descriptive statistics for the model parameters showed mean values of 1.21±0.41 cm (S), 8.25±3.33 %/cm (W), and 708.13±388.64 cc (OBV), while the MHD averaged 3.25±0.78 Gy. The regression model demonstrated a high R² value of 0.9, with significant coefficients for S, W, and OBV. The results suggest that the developed model provides a reliable method for predicting MHD based on treatment geometry, offering potential for better heart dose management in radiotherapy.