Dr. Xiangmin Han, Postdoctoral Fellow, was awarded the Young Scientists Fund of the National Natural Science Foundation of China (Category C)

title: Research on Multidimensional Hypergraph-based Collaborative Computation for Renal Cancer Survival Prediction

Renal cancer, a common malignancy within the urinary system, often exhibits no significant symptoms in its early stages, leading to a majority of patients being diagnosed at advanced stages. Thus, accurate predictions of renal cancer survival are crucial for timely clinical decision-making and the formulation of prognostic treatment plans. This project addresses the current challenges in predicting renal cancer survival based on multimodal medical data, such as the scarcity of effectively annotated data due to labeling difficulties and high costs, and the complex associations caused by diverse and multi-source heterogeneity. By researching a renal cancer survival prediction model based on multimodal hypergraph computations, this project aims to compensate for the lack of high-order associative information in renal cancer. Furthermore, by investigating a model based on multidimensional hypergraph collaborative computations, it seeks to enhance the cross-modal consistency in renal cancer survival predictions. Additionally, by exploring a lightweight renal cancer survival prediction model based on multidimensional hypergraph knowledge distillation, the project aspires to achieve efficient and lightweight predictions for renal cancer survival. This research is expected to overcome key bottlenecks in the current field of renal cancer survival analysis, establish a lightweight and precise survival prediction model from a high-order associative perspective, and address the performance limitations caused by insufficient data. It aims to provide new theoretical and methodological support for personalized clinical diagnosis and treatment, thereby contributing to the development of a Healthy China and an intelligent medical system.

基于多维超图协同计算的肾癌存活预测方法研究

肾癌是一种常见的泌尿系统恶性肿瘤，在早期阶段往往无明显症状，导致许多患者在确诊时已经进展到中晚期。 因此，准确的肾癌存活预测对于及早临床决策和预后治疗方案的制订至关重要。本项目针对当前基于多模态医 学数据的肾癌存活预测方法面临的因标注困难且代价昂贵导致有效标注数据不足和多源异构且交互多样导致的 复杂关联展开研究。通过研究基于多模态超图计算的肾癌存活预测模型，弥补肾癌多模态高阶关联信息缺失问题； 通过研究基于多维超图协同计算的肾癌存活预测模型，提高肾癌存活预测跨模态一致性；通过研究基于多维超图 知识蒸馏的轻量化肾癌存活预测模型，实现轻量化的肾癌高效存活预测。本项目研究将有望突破当前肾癌存 活预测分析领域的关键瓶颈问题，建立高阶关联视角下的轻量化精准存活预测模型，弥补数据不足来带的性能 受限，为临床个性化诊疗提供新的理论和方法支持，助力国家健康中国和智慧医疗体系建立。