MIL vs. Aggregation: Evaluating Patient-Level Survival Prediction Strategies Using Graph-Based Learning

2025-03-29

AI Summary

Effective patient-level representation from multiple whole-slide images (WSIs) remains a critical challenge in cancer survival prediction. Method: This paper systematically compares instance-level modeling, multi-instance learning (MIL)-driven representative slide selection, and various WSI aggregation strategies. We propose a novel framework integrating graph neural network (GNN)-based multi-scale feature extraction, an MIL architecture, and attention-guided slide selection. Contribution/Results: For the first time, we empirically demonstrate MIL’s superiority over naive aggregation or averaging in patient-level prognostic tasks. On the MMIST-ccRCC dataset, our method achieves a +3.2% improvement in C-index and enhanced calibration, confirming that selecting a single representative WSI via attention yields greater discriminative power and clinical interpretability than conventional aggregation. Our core contribution is establishing MIL as the superior paradigm for WSI-based survival prediction and providing a generalizable, attention-driven slide selection framework.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Multi-modal Vision
• Multiagent Systems: Other Foundations of Multi Agent Systems

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Oncologists often rely on a multitude of data, including whole-slide images (WSIs), to guide therapeutic decisions, aiming for the best patient outcome. However, predicting the prognosis of cancer patients can be a challenging task due to tumor heterogeneity and intra-patient variability, and the complexity of analyzing WSIs. These images are extremely large, containing billions of pixels, making direct processing computationally expensive and requiring specialized methods to extract relevant information. Additionally, multiple WSIs from the same patient may capture different tumor regions, some being more informative than others. This raises a fundamental question: Should we use all WSIs to characterize the patient, or should we identify the most representative slide for prognosis? Our work seeks to answer this question by performing a comparison of various strategies for predicting survival at the WSI and patient level. The former treats each WSI as an independent sample, mimicking the strategy adopted in other works, while the latter comprises methods to either aggregate the predictions of the several WSIs or automatically identify the most relevant slide using multiple-instance learning (MIL). Additionally, we evaluate different Graph Neural Networks architectures under these strategies. We conduct our experiments using the MMIST-ccRCC dataset, which comprises patients with clear cell renal cell carcinoma (ccRCC). Our results show that MIL-based selection improves accuracy, suggesting that choosing the most representative slide benefits survival prediction.

Problem

• Evaluates patient-level survival prediction strategies using graph-based learning
• Compares aggregation vs. MIL for selecting informative whole-slide images
• Assesses Graph Neural Networks for cancer prognosis with WSIs

Innovation

• Compares WSI and patient-level survival prediction strategies
• Uses multiple-instance learning for slide selection
• Evaluates Graph Neural Networks architectures