Artificial intelligence-based pathological application to predict regional lymph node metastasis in Papillary Thyroid...

Area of Science:

• Oncology
• Medical Imaging
• Computational Pathology

Background:

• Papillary thyroid cancer (PTC) is the most common type of thyroid malignancy.
• Lymph node metastasis is a critical prognostic factor in PTC.
• Accurate prediction of metastasis aids in treatment planning and patient management.

Purpose of the Study:

• To develop and validate a predictive model for lymph node metastasis in papillary thyroid cancer.
• To leverage pathology image analysis for predicting regional lymph node involvement.
• To make the developed inference algorithm publicly available for further research and validation.

Main Methods:

• Training a predictive model using pathology images from The Cancer Genome Atlas (TCGA) dataset for papillary thyroid cancer.
• Developing a front-end inference model utilizing a center's dataset and graph neural network concepts (probabilistic propagation of nodes).
• Utilizing single pathological images for predicting tumor spread to regional lymph nodes.

Main Results:

• Demonstrated the common occurrence of regional lymph node metastasis in papillary thyroid cancer.
• Showcased the predictability of lymph node metastasis using the developed model.
• Provided a publicly accessible inference algorithm for validation and further study.

Conclusions:

• Regional lymph node metastasis in papillary thyroid cancer is a predictable event.
• The developed model offers a valuable tool for assessing metastasis risk from pathology images.
• The study encourages further research, validation, and data sharing within the scientific community.