From Arno Schaepe’s experience

Automating cancer tissue analysis with patented image mining

Arno is a named inventor on Definiens image-analysis patents and co-authored research on lung tumor analysis, mitosis detection in melanoma and tissue phenomics for cancer immunotherapy.

Industry
Health and life sciences
Organization
Definiens
Arno’s role
Named inventor and co-author

The business need

In 2011 Definiens set out to make pathology quantitative, correlating detailed cell-level data from tissue images with therapy response and patient outcomes to find predictive and prognostic biomarkers.

In melanoma, the iDermatoPath study set out to detect mitoses in H&E-stained tissue sections with software, and its authors found the tool most useful for difficult cases.

What Arno did

  1. Arno is a named co-inventor on Definiens patents for context-driven image mining to generate image-based biomarkers (US 8,594,410, granted 2013, and US 9,245,337, granted 2016).
  2. Arno is also a co-inventor on Definiens patents for a cognition integrator and language (US 7,873,223, granted 2011) and for a user interface for interpreting image-analysis results (US 8,711,149, granted 2014).
  3. Arno co-authored iDermatoPath (2017), a machine-learning tool developed on 59 melanoma specimens that detects tumor regions and mitotic nuclei and ranks them by diagnostic relevance.
  4. Arno co-authored publications on lung tumor analysis and on tissue phenomics for cancer immunotherapy, including a 2019 Annals of Oncology publication on the tumor immune microenvironment.

Technology and approach

  • Object-oriented image analysis and context-driven image mining
  • Machine-learning detection of tumor regions and mitotic nuclei in H&E-stained sections

Outcome

  • iDermatoPath reached a Dice coefficient of 0.72 for tumor regions and more than 83% accuracy in mitosis detection.

Decisions this experience speaks to

  • Should we use classical scientific computing and machine learning, or large language models, for our problem?
  • How do we protect our IP when working with external partners in a sensitive biology space?

Related services and use cases

Sources

  1. Arno Schaepe’s profile on CTO Grand Rounds
  2. US8594410B2 Context driven image mining - patents.google.com (opens in a new tab)
  3. US9245337B2 Context driven image mining - patents.google.com (opens in a new tab)
  4. US7873223B2 Cognition integrator and language - patents.google.com (opens in a new tab)
  5. US8711149B2 Graphical user interface for image analysis results - patents.google.com (opens in a new tab)
  6. Publications by Schäpe A (iDermatoPath, 2017) - Europe PMC (opens in a new tab)
  7. Publications by Arno Schäpe - Semantic Scholar (opens in a new tab)
  8. Definiens releases quantitative digital pathology portfolio (2011) - biospace.com (opens in a new tab)