Table 1 Description of image sets.

From: A machine learning algorithm for simulating immunohistochemistry: development of SOX10 virtual IHC and evaluation on primarily melanocytic neoplasms

Alias

Lesion types

Number of WSIs

Same-layer IHC

Description

NN-master set

MIS, malignant melanoma, MM, BCC, neuroma

12

Yes

All H&E images with corresponding digital IHC masks; see Table 2. The NN-master set was randomly divided into the NN-train set (90%), which was used to train the neural network, and the NN-test set (10%), which was used to test the neural network.

IHC-test set

One case of inflamed melanoma

1

Yes

Image used for graphical evaluation of vIHC with direct comparison to IHC.

Subjective-test set

Invasive melanoma, MM, BCC

6

No

Graphical evaluation of vIHC using H&E only with inferred correct IHC-staining pattern.

  1. BCC basal cell carcinoma, H&E hematoxylin and eosin, IHC immunohistochemistry, MIS melanoma in situ, MM metastatic melanoma, NN neural network, WSI whole slide image.
Back to article page