ESTRO 2024 - Abstract Book

S5019

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

Homogeneity

-0.534

0.262

-2.038

0.042

Multifactorial Joint Modelling Of Study Endpoint I Incorporation of features Estimated coefficients

regression

standard error

Z-value

P-value

Skewness

-0.512

0.234 0.445

-2.191

0.028 0.011

ECOG score

1.134

2.552

Multifactorial Radiomics Modelling for Study Endpoint II Incorporation of features Estimated regression coefficients

standard error

Z-value

P-value

N-staging

1.627

0.589

2.76

0.006

Multifactorial Radiomics/Joint Modelling For Study Endpoint II Incorporation of features Estimated regression coefficients

standard error

Z-value

P-value

Volume

density

-0.479

0.241

-1.992

0.046

(ellipsoid)

Cluster Prominence

0.625

0.276

2.266

0.023

Long run high gray level emphasis

0.722

0.269

2.685

0.007

Conclusion:

In this study, the prediction model of PD-L1 expression in patients with non-small cell lung cancer based on radiomics features and clinical features were constructed. Compared with the traditional pathology slides IHC staining to determine PD-L1 expression, the model provides the advantages of non-invasiveness, real-time, and convenience. In the future, we hope to further combine it with artificial intelligence, use deep learning efficient feature extraction and image recognition methods, and predict the expression of related biomarkers based on pre-treatment images of NSCLC patients, thus providing more efficient and precise technical support for personalized treatment plans for patients.

Keywords: PD-L1 expression level, Immunotherapy

1119

Digital Poster

Radiomics and deep learning for glioma treatment outcome prediction based on tumor invasion modeling

Mehdi Astaraki 1,2 , Wille Häger 1,2 , Marta Lazzeroni 1,2 , Iuliana Toma-Dasu 1,2

1 Stockholm University, Medical Radiation Physics, Stockholm, Sweden. 2 Karolinska Institutet, Oncology-Pathology, Stockholm, Sweden

Purpose/Objective: