Healthcare Big Data for Pathology

Share this Post:

August 27, 2018 – Enabling Healthcare Big Data for Pathology requires moving from single-purpose, single-value pathology data to general-purpose rich information data for tissue.

Manual Microscopy, the current standard of care in Pathology where a pathologist examines a histology slide under a microscope, provides only a single-purpose (specific indication) single diagnosis (e.g. cancer/no cancer) or a single score (e.g. 0, 1+, 2+ or 3+) for an entire histology slide. This is a tremendous reduction in data. A histology slide typically has somewhere between 500,000 to over 1,000,000 cells with complex contextual biology (for example: heterogeneity on a slide) and cellular patterns (for example: tissue context as it relates to Immuno-Oncology) that contain significantly more information.

Pathology AI (Artificial Intelligence) systems need to be designed to provide general-purpose rich information data for tissue on which any number of single-purpose interpretations or scoring schemes can be implemented. Rich information data for tissue consists of cell-type (e.g. tumor, stroma) specific cell distributions with an abundance of measured features. The key to the value of this rich data is the list of biology based features (e.g. nucleus area size, cell density), which still allow for simple interpretation but capture all relevant information on the slide.

What would Healthcare Big Data for Pathology look like?

Imagine a single test in a clinical lab (for any given tissue type) where a standard panel of multiple markers provides consolidated rich information data for the tissue. This general-purpose rich information data would support several specific Diagnostics (Dx), Prognostics (Px) and Companion Diagnostics (CDx) by simply correlating the rich data to clinical outcome, without requiring creating or running a new test.

Treatment decisions, including the full spectrum of all available drugs (and future drugs, see below), could be based on a single test (solving a major problem in Immuno-Oncology). New Diagnostic (Dx), Prognostic (Px) and Companion Diagnostic (CDx) capabilities could be created by clinicians in the field by correlating existing or emerging health conditions with this “live” clinical database.

For drug development, rich information data for tissues would allow better characterization of patient populations, thus creating smarter tissue and patient selection strategies to accelerate and lower the costs of new drug developments.

Rich information data for tissue may also be the key for considerably simplified regulatory pathways. In the traditional setting, every tissue type – stain – clinical outcome combination requires a different FDA approval, a costly and time-consuming endeavor. Basically, every Diagnostic (Dx), Prognostic (Px) and Companion Diagnostic (CDx) test needs its own complex FDA approval. However, it may be possible that a simple FDA clearance per tissue type would be sufficient for rich information data as we are dealing with general measurements which are unrelated to clinical outcome. During drug development, or when using big data in a clinical setting, you would just need to develop a scoring scheme that links the already FDA cleared rich information data to clinical outcome. This would be a much faster and less complex approach.

In the big data realm, traditional single-purpose, single-value pathology data has very limited utility for big data applications. However, general-purpose rich information data allows the full potential of tissue data to be realized in a similar manner as genomics data. General-purpose rich information data for tissue is easy to understand (by a biologist) and easy to access (it’s just a database) using any data science platform, making it fairly strait forward to combine tissue data with other data sources such as Next Generation Sequencing (NGS), another rich information data source complementary to tissue, and radiology.

What is your experience with using pathology data in combination with other data sources, like Next Generation Sequencing (NGS) or Radiology? How would you like to use pathology data in Healthcare Big Data applications?

Flagship Biosciences has been developing our own Pathology AI system over the last 8 years to solve the most challenging real-world tissue image analysis problems across the entire Pharma industry. Our Pathology AI system provides rich information data for tissue and is ready to be used for Healthcare Big Data.

If you want to learn more about rich information data for tissue, check out our previous LinkedIn articles, where we presented a short lecture series about the different aspects about Pathology AI intended for a broad non-technical audience.

Holger Lange, PhD
Chief Technology Officer
Flagship Biosciences

Share This Post: