Artificial intelligence-based pathological model for pan-cancer lymph node metastasis detection: a multicentre...

Area of Science:

• Oncology
• Pathology
• Artificial Intelligence

Background:

• Accurate lymph node metastasis detection is vital for cancer staging and treatment.
• Conventional methods may miss micrometastases, leading to underdiagnosis.
• This study developed an AI model for pan-cancer lymph node metastasis detection.

Purpose of the Study:

• To develop and validate a pan-cancer artificial intelligence diagnostic model (PanCAM) for detecting lymph node metastasis.
• To assess PanCAM's generalisability across common and rare cancer types.
• To compare PanCAM's performance with that of human pathologists.

Main Methods:

• A multicentre study included patients from 17 hospitals in China, encompassing common and rare cancers.
• Whole slide images (WSIs) of lymph nodes were analyzed using supervised and incremental learning strategies.
• Retrospective and prospective validations were performed across multiple institutions, including the CAMELYON16 dataset.

Main Results:

• PanCAM demonstrated high diagnostic sensitivity for lymph node metastasis, ranging from 0.93 to 1.00 in prospective validation.
• The model achieved a sensitivity of 0.98 for rare cancers, despite being trained on common cancer data.
• PanCAM identified additional metastatic cases missed by pathologists in both retrospective and prospective validations.

Conclusions:

• PanCAM offers a generalisable AI solution for detecting lymph node metastasis across diverse cancer types.
• The model's high sensitivity and robust performance can assist pathologists, improving diagnostic accuracy.
• Enhanced diagnosis supports better treatment decisions and potentially improves patient outcomes.