Towards Automated Analysis of Sentinel Lymph Nodes using AI

P. Bándi

• Promotor: B. van Ginneken, G. Litjens and J. van der Laak
• Graduation year: 2024
• Radboud University

Abstract

The outline of the thesis is the following:

Chapter 2 details our study, where we developed convolutional neural networks to distinguish tissue from the background. We compared our CNNs to 3 traditional methods on a range of resolution levels using Dice score and sensitivity. We also tested whether a single neural network can perform equivalently to multiple networks, each specialized in a single resolution.

Chapter 3 describes the CAMELYON dataset composed of 1399 annotated wholeslide images of lymph nodes, both with and without metastases, in total three terabytes of data in the context of the CAMELYON16 and CAMELYON17 Grand Challenges. The slides of this dataset were collected from five different medical centers to cover a broad range of image appearance and staining variations.

Chapter 4 describes the CAMELYON17 challenge. We provided the CAMELYON dataset divided into training and test subsets for the competition. Within the challenge, participants were scored based on the ability of their algorithm to identify the pN-stages of the test patients in the test subset. Additionally, we assessed whether combining algorithms could lead to even better performance than each algorithm individually.

Chapter 5 we investigate how to leverage existing high-quality datasets most efficiently in multi-task settings for closely related tasks. Specifically, we will explore different training and domain adaptation strategies, including the prevention of catastrophic forgetting, for breast, colon, and head-and-neck cancer metastasis detection in lymph nodes.

Chapter 6 provides a general discussion of the work presented in this thesis and an outlook toward the future in this field.