Interpretable Artificial Intelligence across Scales for Next-Generation Cancer Prognostics

Introduction

Computation pathology has the potential to revolutionize cancer care and research, specifically through improving assessment of patient prognosis and treatment selection by applying advanced machine learning methods to digitized tissue sections, i.e. whole-slide images (WSIs). This will allow us to replace the current state-of-the-art of human-developed cancer grading systems.

Challenges in the Field

However, the field is currently hindered by significant knowledge gaps:

1. We do not know how to effectively leverage both global and local information in WSIs.
2. We do not know how to identify pan-cancer prognostic features.
3. We do not know how to make machine learning models explainable and interpretable.

Project Objectives

In this project, I will address these key knowledge gaps by building on the novel stochastic streaming gradient descent (SSGD) developed in my group.

Methodology

Specifically, I will:

• Integrate innovative multi-task and cross-task learning algorithms with SSGD.
• Leverage the latest advances in self-supervision, self-attention, and natural language processing to endow deep neural networks with unprecedented transparency and explainability.

Validation and Impact

Last, the project will validate our developed methodology in the largest dataset of oncological WSIs in the world. For the first time, it will identify links between morphological prognostic features and genetic features.

Scientific Contribution

By publicly releasing all developed tools and data, the proposed project will have a scientific multiplier effect for the fields of oncology, computational pathology, and machine learning.

Applications

Specifically, the derived cancer-specific and pan-cancer biomarkers can be leveraged in:

• Clinical care
• Cancer research

Additionally, the enhanced SSGD method for other tasks in computational pathology and our novel multi-task and explainability algorithms can impact other research areas in machine learning, such as:

• Remote sensing
• Self-driving cars