(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose: Knowledge of tumor growth behavior is key to clinical decision-making in the treatment and follow up management of vestibular schwannoma (VS) patients. In the current clinical practice, a series of images acquired at a fixed interval are used to determine VS tumor growth, which may lead to delayed treatment for fast-growing tumors or over-imaging for slow-growing tumors. In this work, we aim to develop a hybrid predictive model to predict VS tumor growth based on the initial diagnosis imaging of the tumor to aid in personalized management of VS.
Methods: Contrast-enhanced T1-weighted MR images of 65 VS patients from our institute were used in this study. Each patient has two MRI scans with a median interval of 7.8 months. VS on these images were segmented by two observers independently. Thirty-six of these patients had tumor growth, which was defined as an increase above 20% volumetrically between two consecutive scans. The hybrid model used a multi-objective multi-classifier radiomics (MOCR) model and a convolutional neural network (CNN) together to exploit the contextual information from the initial MRI scans for tumor growth prediction. The probability outputs of these two sub-models were then fused through an evidential reasoning (ER) approach to make the final prediction.
Results: Five-fold cross-validation was adopted for the model training, validation and testing. Sensitivity, specificity, accuracy and area under the receiver operating characteristic curve (AUC) of the prediction results on testing samples of the hybrid model are 0.75, 0.76, 0.75 and 0.80, respectively, which are all higher than MOCR- and CNN-based model alone.
Conclusion: We developed a hybrid model for VS tumor growth prediction using the initial diagnosis MRI image. In the hybrid model, the fusion of MOCR and CNN improved the prediction reliability.
Keywords
Image Analysis, Texture Analysis
Taxonomy
IM/TH- Image Analysis (Single Modality or Multi-Modality): Machine learning
Contact Email
