Abstract:
Accurate prediction of glioma grade is significant for treatment planning and management. Prior studies require a segmentation network to extract the tumor region, which was then used by classification network for grade prediction. However, tumor segmentation was a challenging pre-processing task and inaccurate tumor extraction can lead to poor classification performance. In this work, we propose an attention-based model for grade prediction. The model contains attention layers to estimate the regions of interest that are relevant for grade classification. The F1-score of the proposed model is 91.18%, which is at least 6% higher than the state-of-the-art deep learning models. In addition, the proposed model was able to generate a more interpretable output.
Audience Take Away Notes:
The audience will learn about a novel approach to glioma grade prediction that addresses the limitations of previous methods. Specifically, they will understand:
- The significance of accurate glioma grade prediction for treatment planning and management.
- The challenges associated with prior methods, which rely on tumor segmentation followed by classification.
- The proposed solution: an attention-based model for grade prediction that doesn't require explicit tumor segmentation.
- How attention layers are used to identify relevant regions for grade classification, improving accuracy and interpretability.
