ESTRO 2024 - Abstract Book

S2207

Clinical - Upper GI

ESTRO 2024

Radiotherapy has long been a pivotal component in the treatment of esophageal cancer. However, the diverse tumor characteristics and patient anatomies pose challenges in planning optimal treatment. Understanding achievable normal organ doses, particularly for lungs and heart, is crucial for efficient and high-quality planning. This study explores the implementation of auto-planning, considering individual anatomical variations, to enhance planning efficiency and treatment quality.

Material/Methods:

A retrospective analysis of 301 plans was conducted, focusing on 13 anatomical and geometric parameters as potential predictors for heart and lung doses. Stepwise regression analysis identified key parameters, enabling the formulation of a predictive linear regression model. An auto-planning script was developed using Eclipse Scripting API (PyESAPI). The study employed 5-field intensity-modulated radiation therapy (IMRT) with a prescribed dose of 45 Gy in 25 fractions. Evaluation metrics included various parameters such as V 100% , D max , Conformity Index (CI), and Homogeneity Index (HI) for the planning target volume (PTV), as well as heart V 30 and mean dose, and lung V 5 , V 10 , V 20 , and mean dose. For comparison, both auto-plan and manual plan were normalized to PTV V 100 (%) as 95. Paired t-tests compared differences between auto-plans and manual plans.

Results:

The ratio of "Lung Crop PTV volume/lung volume" emerged as the most predictive parameter for lung doses, while the "Heart Crop PTV volume/Heart volume" ratio was the key predictor for heart doses. In the test set, auto-plans achieved comparable PTV coverage, CI, HI, and hotspot values to manual plans, yet demonstrated significantly lower lung and heart doses. Notably, the average time for generating an auto-plan was impressively short, ranging from 3 to 4 minutes, along with reduced monitor units (MU).

Auto-plan

Manual plan

p value

PTV V 100 (%)

95.0 ±0.0

95.0 ±0.0

--

PTV D max (cGy)

108.5 ± 1.4

108.7 ±2.2

0.0440

Lung V 5 (%)

43.3 ± 15.1

45.2 ± 14.6

<0.001

Lung V 10 (%)

27.1 ± 10.0

30.3 ± 10.1

<0.001

Lung V 20 (%)

13.0 ± 3.2

16.0 ± 6.5

<0.001

Lung D mean (cGy)

840.0 ± 284.4

896.1 ± 286.8

<0.001

Heart V 30 (%)

36.7 ± 19.3

36.3 ± 20.7

0.3326

Heart D mean (cGy)

2131.4 ± 933.3

2213.8 ± 992.0

<0.001

MU

712.38 ± 134.1

851.53 ± 262.89

<0.001

Conclusion:

This study identified strong geometric predictors for lung and heart doses. Incorporating these predictors into the planning process resulted in reduced heart and lung doses while maintaining efficiency. Implementation of these findings promises advancements in esophageal cancer radiotherapy, enhancing both patient outcomes and resource utilization.