Predicting Prognosis in Localised Colon Cancer Using Artificial Intelligence

This project is a joint research initiative between UZ Leuven and AZ Delta/RADar Learning & Innovation Center, combining complementary expertise in clinical oncology, pathology, radiology, data science and artificial intelligence. As equal partners, UZ Leuven and AZ Delta jointly lead the development and validation of an AI-driven prognostic model for patients with localised colon cancer.

---

Introduction

Colon cancer is one of the most common cancers worldwide. Although most patients are cured with surgery only, a significant proportion will experience disease recurrence. At time of diagnosis, clinicians currently have limited tools to accurately identify which patients are at highest risk of recurrence. This project aims to address that challenge by developing an artificial intelligence (AI)-based model that combines medical imaging, pathology, and clinical data to predict patient outcomes.

Objective

The primary objective of this project is to develop and validate a multimodal AI model that predicts the patients’ prognosis at time of diagnosis.

Specific goals include:
– Developing AI models based on digital pathology and diagnostic CT imaging.
– Integrating pathology, imaging, and clinical data into a single predictive model.
– Validating the model across multiple healthcare institutions.
– Creating the foundation for a future clinical decision support tool that can help identify patients who may benefit from personalised treatment strategies.

Methodology

The project uses retrospective clinical, imaging, and pathology data from multiple Belgian hospitals. Advanced deep learning techniques are applied to digitised biopsy slides and CT scans to identify patterns associated with disease recurrence.

Separate AI models are developed for pathology and radiology data before being combined into a multimodal prediction model. The resulting model is validated using independent patient cohorts from participating centres to ensure robustness and generalisability.

Impact and future directions

By providing more accurate prognostic information at diagnosis, this project has the potential to improve treatment decision-making for patients with localised colon cancer.

The long-term vision is to develop an accessible clinical decision support system that helps clinicians identify patients who may benefit from a neoadjuvant treatment while avoiding unnecessary treatments for low-risk patients.

General info & contact

Keywords (#): Colon cancer, treatment decision support, artificial intelligence (AI), multimodal learning,
clinical decision support software

RADar project research lead: Dr. Nathalie Mertens

RADar project researchers: Prof. Dr. Peter De Jaeger

Principal investigator: Prof. Dr. Jeroen Dekervel (UZ Leuven)

Site investigators: Dr. Sofie De Meulder (AZ Delta); Dr. Pieter-Jan Cuyle (Imeldaziekenhuis Bonheiden);
Dr. Antoon Billiet (AZ Oostende)

Timeline: 2024-2028

Status: Ongoing

Partners: AZ Delta/RADar Learning & Innovation Center, UZ Leuven, Imeldaziekenhuis Bonheiden, AZ Oostende

Funding: Kom op Tegen Kanker, Roche, Merck, Amgen, Nordic Pharma