Training a classifier to recognize cancer biopsy images.

Question

[Caitlin]

Hi all.

I am planning a semester project for my "Introduction to Computational Biology" course and I am very interested in both cancer and machine learning. I downloaded and installed: Python 3.5.3, keras, theano, tensorflow, matplotlib, mahotas, scikit-learn, and scikit-image. I was wondering if a publicly available dataset of images was available which contained, ideally, several thousand cancer biopsies with which I could train a classifier, i.e., use as a training corpus. Something similar to:

https://www.cs.toronto.edu/~kriz/cifar.html

If so, I would then like to be able to submit images that the classifier as not "seen", so that a determination, non-cancerous, benign, or malignant could be made.

Thank you.

~Caitlin

Answer 1

[Jean-Karim Heriche]

Try the The Cancer Imaging Archive (TCIA).