misas 0.0.3

A library for segmentation model explainability through sensitivity analysis

Model Interpretation through Sensitivity Analysis for Segmentation

Interpret and explain your segmetation models through analysing their sensitivity to defined alterations of the input

Input alterations currently include:

• rotation
• cropping

Install

pip install misas

How to use

Example with kaggle data

from fastai.vision import *
from misas.core import default_cmap

#hide_output
img = lambda: open_image("example/kaggle/images/1-frame014-slice005.png")
trueMask = lambda: open_mask("example/kaggle/masks/1-frame014-slice005.png")
trainedModel = Fastai1_model('chfc-cmi/cmr-seg-tl', 'cmr_seg_base')
fig, ax = plt.subplots(figsize=(8,8))
img().show(ax=ax)
trueMask().show(ax=ax, cmap=default_cmap)

Rotation

plot_series(get_rotation_series(img(), trainedModel))

results = eval_rotation_series(img(), trueMask(), trainedModel)
plt.plot(results['deg'], results['c1'])
plt.plot(results['deg'], results['c2'])
plt.axis([0,360,0,1])

(0.0, 360.0, 0.0, 1.0)

You can use interactive elements to manually explore the impact of rotation

from ipywidgets import interact, interactive, fixed, interact_manual
import ipywidgets as widgets

rotation_series = get_rotation_series(img(),trainedModel,step=10)

def plot_rotation_frame(deg):
    return plot_frame(*rotation_series[int(deg/10)], figsize=(10,10))

#hide_output
interact(
    plot_rotation_frame,
    deg=widgets.IntSlider(min=0, max=360, step=10, value=90, continuous_update=False)
)

There are lots of other transformations to try (e.g. cropping, brightness, contrast, ...) as well as MR specific artifacts.

Overview

This is the schematic overview of how misas works. Created with the amazing Excalidraw.

Logo

The logo was designed by Markus J. Ankenbrand using:

• Open box / Boite ouverte by SimpleIcons via openclipart.org
• Cutter icon by SimpleIcons via openclipart.org, original by Marco Olgio, via WikiMedia
• Hack Font
• Inkscape

Attribution

This project is inspired by the awesome "Is it a Duck or Rabbit" tweet by @minimaxir. Also check out the corresponding repo.

%%html
<blockquote class="twitter-tweet"><p lang="en" dir="ltr">Is it a Duck or a Rabbit? For Google Cloud Vision, it depends how the image is rotated. <a href="https://t.co/a30VzjEXVv">pic.twitter.com/a30VzjEXVv</a></p>&mdash; Max Woolf (@minimaxir) <a href="https://twitter.com/minimaxir/status/1103676561809539072?ref_src=twsrc%5Etfw">March 7, 2019</a></blockquote> <script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script> 

Is it a Duck or a Rabbit? For Google Cloud Vision, it depends how the image is rotated. pic.twitter.com/a30VzjEXVv

— Max Woolf (@minimaxir) March 7, 2019

<script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script>