Metastasis Detection, AI

Developed for metastasis detection in H&E stained lymph nodes

Finding metastases in H&E stained lymph node sections can be time consuming and challenging. The differences between small metastases, epithelial tissue and clusters of macrophages are often subtle, and would not be possible to distinguish using conventional image analysis techniques.

This APP utilizes AI/deep learning and has been trained to detect metastases in lymph nodes associated with breast and colon adenocarcinoma, stained with H&E. The deep learning architecture allows it to recognize complex structures and interpret the tissue context when analyzing an image, making it an efficient tool for detecting even small metastases that are not easily noticed.

RUO

Figure 1

All relevant lymph node tissue is automatically outlined (in purple) for further analysis.

Figure 2

Metastases are identified as High Probability (red), Medium Probability (orange), Low Probability (yellow) and the largest is highlighted and measured.

Figure 3

Zooming in, the largest metastasis is reviewed in detail.

Figure 4

The remaining metastases in the image are reviewed as necessary, showing High Probability (red), Medium Probability (orange) and Low Probability (yellow).

Figure 5

The APP also identifies metastases in lymph nodes associated with colon adenocarcinoma and is robust against H&E stain variations.

Figure 6

The metastases are outlined showing High Probability (red), Medium Probability (orange) and Low Probability (yellow).

Details

Quantitative Output variables

The output variables obtained from this protocol are:

Largest Metastasis Diameter [mm] (for breast)
Total Metastasis Area [mm²] (for colon)

Workflow

The APP contains four protocols:

Step 1: Tissue Detect: Outlines tissue on the slide for further analysis.

Step 2: Metastasis Detection: Identifies possible metastases using AI.

Step 3: Post Processing: Post-processes the classification results, improving accuracy and visualization.

Step 4: Calculate Results: Locates the single largest metastasis and calculates the diameter and/or calculates the total metastasis area.

The protocols can be run separately or as one using the APP sequence functionality in VIS.

Methods

The APP was developed using the using the DeepLabv3+ neural network. The neural network uses a cascade of layers of nonlinear processing units for feature extraction and transformation, with each successive layer using the output from the previous layer as input. DeepLabv3+ uses an encoder-decoder structure with atrous spatial pyramid pooling (ASPP) that is able to encode multi-scale contextual information by probing the incoming features with filters or pooling operations at multiple rates and multiple effective field-of-views. This means that instead of using step-wise upsampling blocks to incorporate features from different levels, this network only needs two upsampling steps, i.e. it is faster to train and analyze than e.g. the U-Net. All of this also means that the decoder module can refine the segmentation results along the object boundaries more precisely. For more information on the network architecture, see [1].

Staining Protocol

There is no staining protocol available.

Additional information

To run the APP, a NVIDIA GPU with minimum 4 GB RAM is required.

Keywords

Lymph node, metastasis, H&E, hematoxylin, eosin, deep learning, AI, image analysis, breast cancer, DeepLabv3+, colon cancer, adenocarcinoma

References

LITERATURE
1. Chen, L., et al. Encoder-decoder with atrous separable convolution for semantic image segmentation. Proceedings of the European conference on computer vision (ECCV) 2018, 801-181, arXiv:1802.02611