To clarify the input modalities and data flow used in survival prediction for colorectal cancer, addressing points raised by Wang et al.
Key Findings:
Survival prediction is based solely on CT data during inference.
Endoscopic data aids in training but is not used during the inference phase.
The model employs a decoupled representation strategy for anatomy and style.
Interpretation:
The findings emphasize the importance of training strategies in multimodal frameworks, highlighting how different data types can enhance model performance without being directly used during inference.
Limitations:
The study does not address potential biases in the training data.
No external validation of the model's performance was discussed.
Conclusion:
The response clarifies the methodology and data flow in the survival prediction model, reinforcing the role of multimodal training in enhancing CT-based predictions.
