Imperial College London Postdoc Dr. Ling Huang Celebrates Major Publication in IEEE Transactions on Fuzzy Systems!

LONDON - Dr. Ling Huang, a postdoctoral research associate at Imperial College London, is making significant waves in the intersection of artificial intelligence and healthcare. Her latest breakthrough paper, titled “EsurvFusion: An evidential multimodal survival fusion model based on Epistemic random fuzzy sets,” has been officially accepted and published in the prestigious journal IEEE Transactions on Fuzzy Systems (January 2026 issue).

The research addresses one of the most complex challenges in modern medical AI: predicting patient survival outcomes by seamlessly integrating vastly different types of medical data.

Tackling the Challenges of Medical Data

In clinical settings, patient data is inherently “multimodal” - meaning doctors rely on a mix of structured clinical records, complex imaging (like MRI or PET scans), free-text clinical notes, and genomic profiles to make decisions. Combining these heterogeneous data sources to predict survival time (a task known as survival analysis) is notoriously difficult.

Furthermore, real-world medical data is often noisy, and the exact survival time of patients is frequently “censored” (partially known due to patients dropping out of a study or the study ending before an event occurs). Traditional machine learning models struggle to effectively navigate this uncertainty.

Enter EsurvFusion

To overcome these hurdles, Dr. Huang and her collaborators developed EsurvFusion, an innovative, interpretable AI model designed to fuse multimodal data at the decision level.

EsurvFusion stands out by utilizing “Epistemic Random Fuzzy Sets” - specifically employing Gaussian random fuzzy numbers (GRFNs) - to meticulously quantify both data uncertainty and model uncertainty. Rather than blindly trusting all data inputs, the model features a unique “reliability discounting layer.” This mechanism acts as a quality control filter, automatically learning the reliability of each data modality and reducing the misleading impact of noisy or low-quality data before making a final prediction.

By gracefully handling missing information and assessing the trustworthiness of each data stream, EsurvFusion provides transparent, highly accurate survival predictions.

Setting a New State-of-the-Art

The team rigorously tested EsurvFusion across four different multimodal cancer survival datasets. The model consistently outperformed existing single-modality and multimodal fusion strategies, establishing a new state-of-the-art benchmark in predictive performance and reliability.

Because the model explicitly highlights the influence of different medical modalities through learned reliability coefficients, it also offers a high degree of interpretability - a crucial factor for clinicians who need to trust and understand AI-driven recommendations.

A Collaborative Global Effort

The research highlights a robust international collaboration. Alongside Dr. Huang, who is jointly affiliated with the Saw Swee Hock School of Public Health at the National University of Singapore (NUS) and the Institute of Clinical Science at Imperial College London, the paper was co-authored by Yucheng Xing, Qika Lin, Jinming Duan, Su Ruan, and Mengling Feng.

Dr. Huang’s ongoing research focuses on developing trustworthy AI solutions, with a special emphasis on uncertainty quantification and reliable decision-making in medical imaging and healthcare. Her latest publication in IEEE Transactions on Fuzzy Systems is a testament to the growing potential of multimodal AI to deliver personalized, data-driven, and life-saving insights in complex medical environments.