intensipy 0.1.0

Normalize intensity values in 3D images.

intensipy

Normalize intensity values in 3D image stacks.

Current Methods

1. Intensify3D

Python implementation of the Intensify3D algorithm originally developed by Yoyan et al. There are some minor adjustments:

1. Semi-quantile normalization is the only Z-normalization method currently implemented.
2. Pixels that are quantile normalized are optionally smoothed using they Savitzky-Galoy method outlined in the original paper. In practice this was necessary to reduce artefact noise.
3. Tissue detection is not currently supported.
4. By default, contrast stretching is performed by skimage.exposure.rescale_intensity(). To perform contrast stretching as implemented by the original Intensify3D, set stretch_method='intensify3d'
5. If no maximum background intensity threshold t is provided, t will be estimated for each slice using Otsu's method.

Original Paper Results

Intensipy Results

Z-normalization Example

Average Intensity Comparison

Installation

Clone the repository and from the terminal run:

pip install .

Example

import numpy as np
import matplotlib.pyplot as plt

from intensipy import Intensify

# decreasing average intensity as z increases.
img_stack = 1 / np.arange(1, 6)[:, np.newaxis, np.newaxis]\
          * np.random.randint(0, 255, (5, 512, 512))                           

for each in img_stack: 
    plt.imshow(each, vmin=img_stack.min(), vmax=img_stack.max(), cmap='gray') 
    plt.show()

model = Intensify()
out = model.normalize(img_stack)

for each in out: 
    plt.imshow(each, vmin=out.min(), vmax=out.max(), cmap='gray') 
    plt.show()

References

1.Yayon, N. et al. Intensify3D: Normalizing signal intensity in large heterogenic image stacks. Scientific Reports 8, 4311 (2018).