Skip to main content

​
Supported Tissue Types

DigiPathAI supports segmentation for three major cancer tissue types, each with specialized model ensembles trained on domain-specific datasets:

Colon Cancer

DigestPath dataset models for colorectal tissue segmentation

Liver Cancer

PAIP dataset models for hepatocellular carcinoma segmentation

Breast Cancer

Camelyon dataset models for metastatic breast cancer detection

​
Model Ensembles

Each tissue type is supported by an ensemble of three state-of-the-art deep learning architectures:

​
Architecture Components

ModelDescriptionStrengths
InceptionV3Multi-scale feature extractionCaptures features at multiple resolutions
DenseNetDense connections between layersEfficient feature reuse and gradient flow
DeepLabV3Atrous spatial pyramid poolingPrecise boundary delineation
When quick=False, DigiPathAI uses all three models in ensemble mode for improved accuracy. When quick=True, only a single model is used for faster inference.

​
Performance Characteristics

​
Inference Modes

Single Model Inference
  • Uses one selected architecture (dense/inception/deeplabv3)
  • Faster processing time
  • Suitable for rapid prototyping
  • Good for resource-constrained environments

​
Model Selection

The tissue type is specified using the mode parameter in the getSegmentation() function: 
from DigiPathAI.Segmentation import getSegmentation

# Colon cancer segmentation
getSegmentation(img_path='slide.tiff', mode='colon')

# Liver cancer segmentation
getSegmentation(img_path='slide.tiff', mode='liver')

# Breast cancer segmentation
getSegmentation(img_path='slide.tiff', mode='breast')
The models are automatically downloaded on first use and cached locally in ~/.DigiPathAI/ for future inference.

​
Next Steps

Explore detailed documentation for each tissue type:

Build docs developers (and LLMs) love

Get started for freeTalk to us