Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

Cross-Modality Masked Learning for Survival Prediction in ICI Treated NSCLC Patients

Created by
  • Haebom

Author

Qilong Xing, Zikai Song, Bingxin Gong, Lian Yang, Junqing Yu, Wei Yang

Outline

This paper presents a large-scale dataset and a novel multimodal feature fusion framework to improve the accuracy of predicting survival in non-small cell lung cancer (NSCLC) patients receiving immune checkpoint inhibitor (ICI) therapy. The large-scale dataset consists of 3D CT images, clinical records, progression-free survival (PFS), and overall survival (OS) data from NSCLC patients. The proposed framework utilizes a cross-modality mask learning approach consisting of two branches, each tailored to a specific modality: a Slice-Depth Transformer for CT images and a Graph-based Transformer for clinical variables. The masked modality learning strategy reconstructs missing components using the intact modality, enhancing the integration of modality-specific features and promoting effective inter-modality relationships and feature interactions. This demonstrates multimodal fusion performance for NSCLC survival prediction that surpasses existing methods and sets a new benchmark.

Takeaways, Limitations

•
Takeaways:
◦
Presenting a large-scale NSCLC patient dataset and a novel multimodal feature fusion framework.
◦
Effective modality feature extraction and fusion using Slice-Depth Transformer and Graph-based Transformer
◦
Improved multimodal integration performance through masked modality learning strategies.
◦
Improving the accuracy of NSCLC survival prediction and suggesting the possibility of establishing personalized treatment plans.
◦
Presenting a new standard for survival prediction models
•
Limitations:
◦
Further validation of the generalizability of the presented dataset is needed.
◦
Research is needed to investigate the applicability of the framework to other types of cancer or treatments.
◦
Further research is needed on the model's interpretability and explainability.
◦
Analysis of dataset bias and potential errors is needed.
👍
Made with Slashpage