Inez Verpalen

Amsterdam UMC

In short, Inez Verpalen's research entails the following:

Development and implementation of Image-based machine-learning models to Determine the response to neoadjuvant therApy in panCreatic ducTal adenocarcinoma (DIDACT)

Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal form of cancer, and surgery is the only curative option. Neoadjuvant therapy (NAT) is administered before surgery for local tumor control and to target micrometastasis. Despite NAT, 40% of patients experience early diseases recurrence (within 12 months post-surgery), resulting in poor survival. Currently, there, is no reliable preoperative biomarker to predict these unfavorable survival outcomes.

In this project, we aim to develop a model using artificial intelligence (AI) to predict NAT response in PDAC patients based on computed tomography (CT) imaging. CT imaging may contain relevant information not visible to the human eye. By using AI, we hope to identify therapy-induced changes and to predict survival based on CT data. We will develop and validate this model using multicenter data from the Netherlands and international data from our consortium PHAIR-consortium partners. Applying this model is expected to lead to improved patient selection for surgery and chemotherapy, ultimately enhancing the quality of life for PDAC patients.

Short summary of progress/results January 2026

The study, which officially commenced in February 2024, continues to advance towards its aim of developing and externally validating CT-based machine-learning models for pre-operative response assessment to neoadjuvant therapy (NAT) in PDAC patients. Over the past year, significant progress has been made across all work packages.

External collaborating centers (PHAIR-consortium) are currently engaged in local data collection as part of WP3, with several sites having already completed or nearing completion of their datasets. In parallel, research activities have focused on data collection, the extraction of CT features and their use for NAT response prediction, in accordance with WP1 and WP2. As part of WP2, we are conducting a feasibility study on micro-CT imaging of pancreatic tissue. This study aims to explore the potential of micro-CT for radiological-pathological correlation in resected pancreatic cancer specimens, with the goal of improving our understanding of tumor characteristics and therapy response.

Additionally, we are in the process of establishing a national PDAC imaging database in close collaboration with the Dutch Pancreatic Cancer Group (DPCG) and Eindhoven (Catharina Hospital), as outlined in WP4. Preparatory steps for database implementation have been taken, and coordination with national partners is ongoing to ensure a robust and sustainable infrastructure. Notably, the recent KWF grant awarded to Catharina Hospital includes dedicated funding for the Collective Minds platform, which has helped to resolve part of our national funding challenges for imaging infrastructure, an issue for which we previously submitted a request to Hanarth, but which was not granted.  Furthermore, we have presented our imaging infrastructure plans to the DPCG (the presentation is included as an attachment to this report).

Overall, the project remains on track, with steady progress in both national and international components, and the team is optimistic about achieving the project objectives within the planned timeline.

Short summary of progress/results January 2025

The study officially started in late February 2024 and aims to develop and externally validate CT-based machine-learning models for pre-operative response assessment to neoadjuvant therapy (NAT) in PDAC patients. Nationally, multiple centers are participating, with approval for retrospective data use from prior studies granted by the Dutch Pancreatic Group (DPCG). Data and imaging collection are progressing as planned. A segmentation model was build, and a preparatory study on a small dataset has been conducted to compare two segmentation methods and handcrafted radiomics versus deep learning approaches.

The next phase will focus on collecting international data through the Pancreatobiliary and Hepatic Artificial Intelligence Research consortium (PHAIR), for which some collaborating centers have already obtained ethical approval.

Additionally, we are exploring the creation of a dedicated PDAC imaging database. Initial discussions with a potential industrial partner are promising and may lead to its successful implementation.