Tissue Detection¶

Tissue detection identifies regions containing tissue vs. background in whole slide images. This is a critical preprocessing step in digital pathology pipelines.

Important

The built-in OtsuTissueDetector is a reference implementation meant to demonstrate the TissueDetector interface. For production use, we strongly recommend implementing your own detector tailored to your specific tissue types, staining protocols, and scanning artefacts.

Tissue detection mask — Otsu thresholding on heart tissue — original region, binary tissue mask, and overlay¶

OtsuTissueDetector¶

from glasscut import OtsuTissueDetector
from PIL import Image

detector = OtsuTissueDetector()
image = Image.open("region.png")
mask = detector.detect(image)  # np.ndarray, 0=background, 1=tissue

The detector converts the image to grayscale, applies Otsu’s thresholding via skimage.filters.threshold_otsu, and returns a binary mask.

Custom Detectors¶

Implement a custom tissue detector by subclassing TissueDetector:

from glasscut.tissue_detectors import TissueDetector
import numpy as np
from PIL import Image

class MyTissueDetector(TissueDetector):
   def detect(self, image: Image.Image) -> np.ndarray:
      # Your custom detection logic here
      # Return binary mask: 0 = background, 1 = tissue
      pass

Using with GridTiler¶

Pass your detector instance to the tiler:

tiler = GridTiler(
   tissue_detector=MyTissueDetector(),
   min_tissue_ratio=0.2,
)