Aim

Lymph node assessment is an essential part of cancer staging. Pathologists must determine both the presence of lymph node metastases and the number of lymph nodes examined, as both contribute to N staging and may influence treatment decisions. Adequate staging often requires examination of many lymph nodes, and individual nodes may be divided into multiple tissue fragments during processing. As a result, a pathologist may need to review dozens of slides for a single patient, making N staging a repetitive and time-consuming task.

Our aim is to build on our lymph node metastasis detection model, MetAssist 2.0, by extending metastasis detection to a broader range of cancer types and developing an image-based system for automated lymph node counting. For this. A key challenge is that one lymph node may appear as several fragments or across multiple sections. For a reliable lymph node count, we aim to combine learned image representations with morphological information and integrate evidence across fragments.

Ultimately, the project aims to combine automated lymph node counting with metastasis detection into a single computational workflow for an efficient and reproducible AI-assisted N staging across cancer types.