IHC Cell Detection App explained

The IHC Cell Detection App features a universal subcellular IHC cell AI that is compatible with a wide range of markers, tissues, and cell types. Our 23-minute video tutorial guides MIKAIA^® step by step through the app.

Time codes of chapters

00:00 Understanding Basic Parameters 02:37 Tissue Detection 03:38 Understanding AI Models AI vs. XXL AI 04:59 Cell Detection: Stain Unmixing & Cell Size 06:57 Classifying Cells as Positive or Negative 10:09 Using Filters for “False Positives” and “Ignore” Class 13:10 Computing DAB Expression 17:05 Grading of Cells & H-Score 19:58 Postprocessing Options: Hotspots 21:38 Postprocessing Options: Clusters

Introducing the IHC Cell Detection App

Contents covered in this first part of the tutorial:

• Learn how to use basic parameters.
• Start-analysis buttons
  • Analyze region of interest (RoI), field of view (FoV), slide, and batch.
  • Re-analyze spreviously selected regions.
• Learn how to view the measured attributes of specific cells.
• Understand the differences between the two AI models — “AI” and “XXL AI” — and when to use which.
• Stain unmixing
• Cell filtering options (e.g., filter by size)
• Configuring settings to determine when a cell is classified as positive or negative, including the “auto” threshold mode.
• Reading and using diagrams with focus on scatter plots.
• Using filters for “false positives” and the “ignore” class.

Classifying membranous or cytoplasmic stains

Contents covered in this second part of the tutorial (starting at minute 13:10):

• Problem: Classifying membrane stains based on average DAB expression across the entire cell is not a reliable foundation for determining positivity and grading cells with membranous or cytoplasmic stains.
  • Solution 1: Estimate the nucleus by shrinking detected cell contours and determine the average DAB expression separately for both cell compartments. Base positivity decisions solely on expression in the outer compartment.
  • Solution 2: Instead of measuring average intensity, focus on the average intensity of the most intensely stained subregion (e.g., the top 20%).
  • Solution 3: Combine of solutions 1 and 2.
• Compute H-score: Enable grading of cells into low, moderate, or strong positive categories.
• Optional postprocessing step: Find hotspots.
• Optional postprocessing step: Group nearby positive cells into clusters and obtain clustering statistics