ESTRO 2024 - Abstract Book

S4978

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

Median follow-up was 37 months. 153 patients were included. Among them, 51 presented a G ≥ 2 RP (33.3% of whom 28.8% G2 RP, 3.2% G3 RP, 1.3% G5 RP). Median age was 65 years, 75.8% of included patients were men, 51.9% had chronic obstructive pulmonary disease. Most of patients received concurrent chemotherapy (92.2%), 3DCRT (57.5% versus 42.5% of IMRT) and a normofractionated treatment (88.2%) with a median total dose of 66 Gy. Clinical and dosimetric variables significantly associated with G ≥ 2 RP included a high initial hemoglobin level (p=0.003), prior cancer, mean dose to healthy lungs (p=0.006), the “lungs-PTV” volume receiving more than 20 Gy: V20Gy (p=0.023) and 13 Gy: V13Gy (p=0.005), and the addition of maintenance durvalumab (p=0.026). Two hundred and eighty radiomic features were extracted from the initial CT scans of whom 95 were finally selected and partitioned into 7 subgroups by Agglomerative Hierarchical Clustering. Seven radiomics features were selected : 4 first order features with a high value associated with the occurrence of G ≥ 2 RP : mean (IBSI Q4LE), variation coefficient (IBSI 7TET), first percentile, and variance (IBSI CH89), and 3 textural features: NGTDM_Contrast (IBSI 65HE), GLDZM_SDLGLE (IBSI RUVG), and GLCM_Normalized_Inverse_Difference (IBSI NDRX).

In our explanatory model, other factors were associated with an increased risk of G ≥ 2 RP such as older age, low Tiffeneau ratio and a decreased initial platelet count.

The best-performing predictive model was the random forest learning model using clinical, dosimetric and radiomic variables, with an area under the ROC curve of 0.72 (95% CI [0.63; 0.80]) versus 0.64 for models using one type of data.

Conclusion:

The addition of radiomics features to clinical and dosimetric ones improves prediction of the occurrence of G ≥ 2 RP in patients receiving thoracic radiotherapy for lung cancer. Identification of a high risk of RP subgroup of patients may lead clinicians to apply stricter dosimetric constraints, intensify their follow-up or discuss medical RP treatment early on.

Keywords: radiomics, radiation pneumonitis, lung cancer

580

Digital Poster

Automatic quantification of metastatic melanoma lesion heterogeneity to prioritise radiotherapy

Mikaela Dell'oro 1,2 , Roslyn J Francis 3,4 , Elin S Gray 5 , Martin A Ebert 6,1,7 , Michael Millward 3

1 The University of Western Australia, Australian Centre for Quantitative Imaging, School of Medicine, Perth, Australia. 2 Harry Perkins Institute of Medical Research, QEII, Perth, Australia. 3 The University of Western Australia, School of Medicine, Perth, Australia. 4 Sir Charles Gairdner Hospital, Department of Nuclear Medicine, Perth, Australia. 5 Edith Cowan University, Centre for Precision Health and School of Medical and Health Sciences, Perth, Australia. 6 Sir Charles Gairdner Hospital, Department of Radiation Oncology, Perth, Australia. 7 The University of Western Australia, School of Physics, Mathematics and Computing, Perth, Australia