Edit model card

(Skin) Melanoma

This model can additionally be run on our pathology reports platform

Credits: Dr. Alberto Leon and Dr. Phedias Diamandis (University of Toronto, Canada)

Introduction

This H&E skin cutaneous melanoma tissue classifier was developed using transfer learning on a histology optimized version of the VGG19 CNN (DOI: 10.1038/s42256-019-0068-6) and trained to recognize skin cutaneous melanoma and other surrounding tissue elements.

Annotations were carried out on batches of image tiles (dimensions: 256 x 256 px) grouped using image-based clustering (HAVOC, DOI: 10.1126/sciadv.adg1894) from 7 TCGA H&E-stained whole slide images. Validation testing was carried out on non-overlapping cases from TCGA.

Classes

  1. Adipose Tissue
  2. Blank space
  3. Blood
  4. Connective_fibrodsis
  5. Intestinal Mucosa
  6. Lymph_dense
  7. Melanoma
  8. Necrosis
  9. Skin Epithelium

This information can be found in the inference.json file

Evaluation Metrics

Classifier validation can be found on the pathology reports platform

Downloads last month
10
Inference Examples
Inference API (serverless) does not yet support tf-keras models for this pipeline type.