ESTRO meets Asia 2024 - Abstract Book

S351

Physics – Radiomics, functional and biological imaging and outcome prediction

ESTRO meets Asia 2024

We estimated breast composition by categorising each voxel of the radiotherapy planning computed tomography (CT) into “ fat ” and “ fibroglandular ” based on intensity. A Gaussian Mixture Model (assuming two Gaussians) was applied to the CT intensities of the contralateral and ipsilateral breasts to characterise individual patient-specific fat and fibroglandular via their mean CT intensity (µ) and standard deviation (σ). We used 2σ from the mean value of individual patients to create fat and fibroglandular breast sub-structures. Dosimetric parameters (mean and max dose, dose uniformity) were extracted for each sub-structure and compared for each severity level of breast oedema using a paired t-test. The correlation between acute breast oedema, dosimetric parameters in sub structures and whole breast contour, patient and clinical variables was investigated using univariable and multivariable ordinal regression analysis. The fit of the model was evaluated using the Akaike and Bayesian Information Criterion (AIC and BIC) (SPSS v29). 13% of patients experienced grade 2 acute breast oedema. CT intensity from the contralateral breast was highly correlated with the ipsilateral breast ( R 2 >0.8). The average of µ and σ were -109 and 7 for fat and -27 and 33 for fibroglandular in the ipsilateral breast. Significantly higher mean dose and dose uniformity were observed for the fibroglandular sub-structure compared to fat at all severity levels ( p <0.05), while no significant difference was found for the maximum dose. The maximum difference in mean dose between fat and fibroglandular sub structure was observed at grade 3 toxicity level (1.14±1.86 Gy). In univariable analysis, body mass index, breast cup size, breast volume, fraction size, and all dose metrics extracted from sub-structure contours were significantly correlated with acute oedema ( p <0.05). Breast volume and dose uniformity remained significant in all multivariable models. The mean dose in fat and fibroglandular substructures was significantly correlated with breast oedema when including only the mean dose parameter in the model. In addition, only the mean dose to fibroglandular was significant in the mean dose plus dose uniformity model. The best-performing model was the model including all dosimetric parameters to the fat sub-structure (AIC=1900, BIC=1953), which was 26 and 36 points lower than the baseline model (excluding dose parameter) and 2 points lower than the model including dosimetric from the whole breast contour. Dose uniformity in contour is the most influential parameter when generating a model. Results:

Conclusion:

Assessing dose to fibroglandular and fat within the breast is feasible. In this cohort, we observed a trend towards higher dose deposition in fibroglandular tissue versus fat. This dose difference was more pronounced in patients experiencing a higher grade of breast oedema. If confirmed in larger cohorts, these findings could help predict and possibly prevent breast oedema after radiotherapy.

Keywords: Breast Composition, Oedema, Breast Radiotherapy