ESTRO meets Asia 2024 - Abstract Book

S116

Interdisciplinary – Education in radiation oncology

ESTRO meets Asia 2024

1. Rosenblatt E, Prajogi GB, Barton M, Fidarova E, Eriksen JG, Haffty B, et al. Need for competency-based radiation oncology education in developing countries. Creative Education. 2017;8(1):66-80.

2. Clinical Oncology Postgraduate Training in Malaysia: Medical Deans Council of Malaysia; 2021.

3. Lim GCC. Clinical oncology in Malaysia: 1914 to present. Biomed Imaging Interv J. 2006;2(1):e18.

4. Postgraduate Handbook. Universiti Malaya; 2022/2023.

14

Proffered Paper

MRI-BASED RADIOMICS MODEL PREDICTS PROGRESSION AND PSEUDOPROGRESSION IN HIGH-GRADE GLIOMA

Ningping Xiao, Yi Jin, pei yang, Rongyao Fang, Wenhui Lu, Jiayu Xiang, Tingjie Yu, Zhengda Pei, Yihui Zhang

Department of Radiation Oncology, Hunan Cancer Hospital, changsha, China

Purpose/Objective:

Ture progression (TP) is characterized by actual tumor growth [1] , invasion of surrounding tissues, and development of new active tumor areas over time, while pseudoprogression (PsP) [2-4] refers to clinical and imaging changes that resemble TP in patients with high-grade glioma (HGG), typically occurring within the first 6 months [5] after surgery and temozolomide concurrent chemoradiotherapy (CCRT). Differentiating between TP and PsP accurately after standard treatment poses significant challenges for radiologists and clinicians. This distinction is crucial as it guides clinical decision-making and impacts patient outcomes [6] . Therefore, we used machine learning (ML) [7,8] to develop and validate the radiomics model from MRI [9,10] to distinguish PsP from tumor progression in patients with HGG.

Material/Methods:

This retrospective study enrolled HGG patients who exhibit progressive enlargement or new enhancement within the radiation field on MRI within 6 months of receiving standard treatment. The region of interest (ROI) for the tumors was delineated using the 3D Slicer software (version 5.0.2) on MR images.Radiomics features were extracted from raw images and wavelet-transformed images based on machine learning methods. The least absolute shrinkage and selection operator (LASSO) method was employed to select discriminative features to reduce the number of features to optimize classification. Using the selected features, support vector machine (Support Vector Machine, SVM) was applied to build a classifier, To evaluate the performance of the model, 10 times 5-fold cross-validation and calculated the mean and variance of the 10 experimental results were performed. The diagnostic performance of the radiomics model was assessed using metrics such as the area under the ROC curve (AUC), accuracy, sensitivity, F1-score, and specificity.

Results:

61 HGG patients who had new or enlarged enhancement within the radiation field on MRI within 6 months after standard therapy, including 42 TP and 19 PsP, were enrolled. Based on the T1-weighted 、 T2-weighted and T1-