July 15, 2018 – Over the past few years, deep learning has created quite a hype about artificial intelligence (AI), and healthcare AI has become a hot topic. Advancements in AI technology for digital pathology may unravel some of the most challenging real-world tissue analysis problems in pharmaceutical development.
To provide a better understanding of the different facets around pathology AI, we created a short video lecture series for a broad, non-technical audience.
This series will explain why end-to-end deep learning is NOT the way to go, and why the true barrier for pathology AI is the business model, not the technology.
- Digitization is Revolutionizing Medicine: We first look at another discipline in medicine that already went digital: radiology. Next, we examine an application of artificial intelligence that can affect human lives and health as well, autonomous driving, to see if there are lessons that can be applied to pathology.
- Artificial Intelligence (AI): To provide a good foundation to understand pathology AI this video provides a short primer on artificial intelligence and machine learning.
- Pathology to Pathology AI (machine learning): We introduce pathology AI step-by-step, starting with an introduction to pathology, going to digital pathology and then adding machine learning to create a pathology AI system. We are going to present a smart system design with the pathologist playing a vital role in the center of the pathology AI system and a patient-type based machine learning approach that allows to overcome a key problem in pathology AI, the variations between patients. To illustrate our concepts, we will provide a demo of one pathology AI system.
- Big Data: A key concept we are going to present is that a pathology AI system needs to provide rich information data for tissue to enable healthcare big data for pathology, allowing us to benefit from the power of data science.
- Regulatory Affairs: We introduce the regulatory landscape, which can be quite complex, reaching from a central lab model under CLIA or CAP regulations to a medical device development under FDA regulations. We then discuss how rich information data for tissue can be the key to considerably simplify the regulatory pathway.
- Business Case: As part of the business case for pathology AI we talk about the market, the required competencies of the companies that want to play in this space and discuss what the killer app and a viable business model would look like.
Holger Lange, PhD
Chief Technology Officer