Foundation models for precision non-small cell lung cancer treatment
Principal Investigator
Name
Mitchell Chen
Degrees
BMBCh,MEng, DPhil, FRCR
Institution
Imperial College London
Position Title
MRC Clinician Scientist
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-1490
Initial CDAS Request Approval
Jan 26, 2026
Title
Foundation models for precision non-small cell lung cancer treatment
Summary
Precision oncology has transformed the way cancer is treated, moving from “one-size-fits-all” protocols toward individualised strategies. Non-small cell lung cancer (NSCLC) is the most common cause of cancer-related mortality worldwide. While newer treatment options such as checkpoint blockade immunotherapy (CBI) have significantly improved survival in a subset of its patients, identifying such patients remains difficult due to the imperfect treatment-guiding methods currently in use. Despite various attempts to use AI on imaging data to find an alternative non-invasive solution to tackle this challenge, most computational tools developed for NSCLC do not fully integrate multimodal data and therefore often fall short in their biological explainability, hindering clinical acceptance and deployment.
Foundation models, large-scale machine learning systems pretrained on multimodal biomedical data, offer a promising paradigm to bridge this gap 3, with the potential to guide real-time clinical decision-making, such as selecting patients for CBI versus alternative treatments, and identifying those at risk of early relapse who may benefit from closer monitoring.
Foundation models are at the forefront of artificial intelligence in precision medicine, offering powerful tools to learn generalisable patterns from large, heterogeneous datasets. Given their ability to capture complex biological signals across diseases and modalities, we propose to explore and apply this concept to our study, by testing the cross-modal knowledge gained from our works on NSCLC on a different malignancy.
Aims
To develop an imaging data-driven foundation model for non-small cell lung cancer (NSCLC).
To train the foundation model to undertake individual tasks informing contemporary precision oncology practice, including patient prognostication, precision treatment planning (in immunotherapy, targeted therapy and tumour ablation) and treatment-related adverse event prediction.
To validate the foundation model on multi-modal data-driven tasks incorporating clinical, histopathological and molecular data alongside imaging, to improve its performance in specific use cases.
To identify robust biomarkers of therapeutic response, resistance, and prognosis through self-supervised representation learning.
To ensure model interpretability using attention mechanisms, causal inference, and domain-informed constraints, reinforcing the model with explanability.
Collaborators
Mitchell Chen Imperial College London
Jillian Yong Xin Sieh Imperial College London
Meng Zhou Imperial College London
Charlie Johnson Imperial College London