Click here to

Session: Outcome Modeling and Assessment [Return to Session]

MRI-Based Multi-Region Delta-Radiomics for Prediction of Recurrence After Chemoradiotherapy in Locally Advanced Cervical Cancer

M Kawamura1*, D Kawahara2, I Nishibuchi2, A Saito2, T Yoshida3, Y Nagata2,4, (1)Graduate School of Biomedical & Health Sciences, Hiroshima University, Hiroshima, JP, (2)Department of Radiation Oncology, Institute of Biomedical & Health Sciences, Hiroshima University, Hiroshima, JP, (3)School of Medicine, Hiroshima University, Hiroshima, JP, (4)Hiroshima High Precision Radiotherapy Cancer Center, Hiroshima, JP


TU-A-TRACK 6-3 (Tuesday, 7/27/2021) 10:30 AM - 11:30 AM [Eastern Time (GMT-4)]

Purpose: The aim of this study is to predict the recurrence of locally advanced cervical cancer after radiotherapy from delta-radiomics features using magnetic resonance imaging (MRI).

Methods: This study included 61 patients with locally advanced cervical cancer treated with chemoradiotherapy between 2003 and 2017. Region of interest(ROI) were created with the uterus as the clinical target volume(CTV). In addition to the ROIs that expanded and shrinked from CTV were created, the inner and shell regions were analyzed. The inner region was defined as the region that the voxels excluded the tumor boundaries. The shell region was defined as the region that the voxels around the tumor boundaries were analyzed. Delta-radiomics feature was defined as the difference in radiomics feature before and at 4 weeks after radiation therapy begins. A total of 9394 features were extracted from T1-weighted images(T1WI) and T2-weighted images(T2WI). Using the features selected by the method of least absolute shrinkage and selection operator (LASSO) regression, two single sequence radiomics models and mixture of the two sequence were constructed using a neural network with 5-fold cross-validation. The performance of the prediction models was assessed in terms of sensitivity, specificity, accuracy and area under the receiver operator characteristic curve (AUC).

Results: From Lasso regression, 4 radiomics features from T1WI and 8 radiomics features from T2WI and 8 radiomics features from combination of T1WI and T2WI. The accuracy, specificity, sensitivity, and AUC of the prediction model for the dataset were 70.0%, 87.5%, 41.2%, and 0.77 with T1WI, 85.6%, 91.9%, 71.4%, and 0.90 with T2WI, 82.2%, 87.8%, 75.6%, and 0.91 with the combined.

Conclusion: The current study showed that multi-region delta-radiomics features could be an important factor to predict recurrence of cervical cancer.



    Not Applicable / None Entered.


    Not Applicable / None Entered.

    Contact Email