# OVH‐guided planning for superior heart and lung sparing in breast cancer radiotherapy

## TL;DR

This study introduces an automated workflow using overlap volume histograms to improve heart and lung protection in breast cancer radiotherapy, especially for post-mastectomy cases.

## Contribution

A novel automated planning workflow using OVH to predict DVH constraints for improved cardiopulmonary sparing in breast cancer radiotherapy.

## Key findings

- Automated plans significantly reduced heart and lung doses in post-mastectomy radiotherapy compared to manual plans.
- The workflow maintained target coverage and dose homogeneity comparable to manual plans.
- Linear correlations between OVH metrics and dose constraints were strong and statistically significant.

## Abstract

Manual planning in breast cancer radiotherapy is often time‐consuming and operator‐dependent, leading to inconsistencies in plan quality. This study validated an automated workflow using overlap volume histograms (OVH) to predict patient‐specific dose–volume histogram (DVH) constraints, aiming to enhance cardiopulmonary sparing and planning efficiency.

A historical database of 322 patients was stratified into four groups: left/right post‐mastectomy radiotherapy (PMRMRT) and left/right breast‐conserving radiotherapy (BCRT). Linear regression models were established to correlate OVH‐derived geometric metrics (Lx
) with corresponding DVH‐based dose constraints (Dx
). These predictive models were integrated into the Monaco treatment planning system via a custom Python script to provide an improved automated planning workflow. The workflow's performance was prospectively validated on 80 independent testing cases (20 per group). Automated plans were generated using the predicted constraints and compared dosimetrically against clinically approved manual plans.

Significant linear correlations were observed between Lx
and Dx
for all OARs (r2
 = 0.51–0.72, p < 0.001). In the PMRMRT testing cohorts, the automated workflow significantly reduced doses to the heart and ipsilateral lung compared to manual planning (p < 0.05). For left‐sided PMRMRT, the heart dose was reduced by 15.6% (D
10), 18.7% (D
5), and 9.8% (D
mean), while the ipsilateral lung dose decreased by up to 6.3% (D
mean). In BCRT cases, automated plans were not significant improved compared to manual plans. Importantly, all automated plans maintained target volume coverage and dose homogeneity comparable to manual plans (p > 0.05).

The OVH‐based framework effectively translated anatomy into achievable objectives, significantly improving heart and lung sparing for complex PMRMRT cases while streamlining clinical workflows.