Akademik Veri Yönetim Sistemi
European Congress of Radiology 2025, Vienna, Avusturya, 26 Şubat - 02 Mart 2025, ss.149, (Tam Metin Bildiri)
Purpose or Learning Objective: To evaluate the potential of machine learning-based models for predicting the response of large hepatocellular carcinoma to transarterial radioembolization Methods or Background: A total of 49 patients (38 responder and 11 nonresponder) were included in the study. Laboratory results and clinical conditions were collected. Treatment response was assessed according to mRECIST criteria from the 3-month follow-up MR examinations. Complete or partial response was categorized as the responder group, while stable or progressive disease was classified as the non-responder group. Radiomics features were extracted from contrast-enhanced T1-weighted images (CE-T1) and T2-weighted images (T2WI). 141 radiomics features were obtained from each lesion. Classification learning models were used to create prediction models for TARE response. 5-fold cross-validation technique was utilized to identify the prediction rates of treatment response. Results or Findings: Number of radiomics features demonstrated statistically significant differences between the groups are 9 and 12 on T2W and CE-T1 images, respectively. The model based on radiomics features obtained from CE-T1 images demonstrated an accuracy rate of %79.6 to predict response with an AUC of 0.92. The sensitivity and specificity rates were %79 and %100, respectively. The accuracy and AUC rates of the model using radiomics features extracted from T2W images were %79.6 and 0.77, respectively. Sensitivity and specificity rates of the model were %80 and %67, respectively. When only clinical and laboratory parameters were used, the model showed an accuracy rate of %77.6 and an AUC of 0.65. The sensitivity and specificity values of the clinical and laboratory model were %79 and %50, respectively. Another model using both clinical and CE-T1 radiomics features showed an accuracy rate of %73.5 Conclusion: Machine learning-based radiomics models based on MRI can predict the response of large hepatocellular carcinoma to transarterial radioembolization