Loading...

Graph Neural Networks Interpretability Diagnosis using Histopathological Images

Abdous, Sina | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57441 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Rohban, Mohammad Hossein
  7. Abstract:
  8. Deep learning methods are rapidly gaining traction for clinical use in digital pathology. Despite the increasing use of graph neural networks for classifying histopathology images, due to the high accuracy of these methods in this field and the introduction of new interpretability methods for these networks, current proposed solutions still face two main issues. First, there is no comprehensive framework for evaluating the effectiveness of interpretability methods for graph networks, particularly for histopathology images. Additionally, applying conventional interpretability methods to these types of networks for pathology images, with consideration of domain-specific knowledge, has been discussed and criticized in recent years. In this project, taking into account datasets of histopathology images from various tissues and prioritizing the difficulty of their classification, different interpretability methods for these networks were examined, and the effectiveness and usefulness of existing methods were evaluated. This evaluation was based on the results obtained from comparing current existing metrics. One criterion for the usefulness of interpretability metrics is the ability to provide a simple and effective explanation to experts and also the ability to validate and debug the predictions of graph neural networks for a specific input sample. Based on this, a new interpretability method was developed that offers better quality according to common metrics in this field and achives the greatest value for Fidelity (0.75 on BRACS and 0.71 on BACH datasets) which is one of the most important metrics in this field. This project utilized publicly available datasets, such as BRACS and BACH in the field of breast cancer, where diagnosing the disease stage is challenging
  9. Keywords:
  10. Graph Neural Network ; Interpretability ; Disease Diagnosis ; Histopathology Images ; Deep Learning

 Digital Object List

 Bookmark

...see more