One map — the red dots are lymphocytes

Shave a thin slice of cancer tissue onto a glass slide, stain it with haematoxylin and eosin (H&E), and the nuclei come out purple and the cytoplasm pink. A pathologist looks at that slide down a microscope and makes a diagnosis.

What is in the picture is not only cancer cells. Lymphocytes — the immune cells the body sent to attack the tumour — are in there too. Small, dark, round nuclei packed close together: that is them.

How many of those lymphocytes have arrived has been an important marker for a long time. Tumours with more of them have been reported to do better, in several cancer types. But how do you count "how many"?

The map has only three colours.

So this file is not a photograph. It is a table of verdicts painted in colour. One pixel corresponds to one 50 µm tile of tissue.

Turning the three colours into 0 · 1 · 2 makes them easier to handle. That is all the read_map() of the til package we use throughout this course does.

def read_map(path):
    a = np.array(Image.open(path).convert("RGB"))
    R, G, B = a[:, :, 0], a[:, :, 1], a[:, :, 2]
    code = np.zeros(a.shape[:2], np.uint8)
    code[(R == 0)   & (G == 0) & (B == 255)] = 1   # tissue
    code[(R == 255) & (G == 0) & (B == 0)  ] = 2   # lymphocytes
    return code