Coder Social home page Coder Social logo

pypatchmatch's Introduction

pypatchmatch

"Once upon a time
You randomly croped some tiles
From some whole slides
Where does the tile lie in slide
Pixel-wise comparison will let you die
Thank God that here is a light
`pypatchmatch` will save your life"

pypatchmatch is a python implementation that can help you efficiently find the location of all patches in one of several WSIs (whole slide images).

Table of Contents

Installation

  1. Clone or download

    • Use the command bellow in terminal to clone this repo:
    git clone https://github.com/samson6460/pypatchmatch.git
    
    • Or just download whole files using the [Code > Download ZIP] button in the upper right corner.
  2. Install dependent packages:

    cd pypatchmatch
    pip install -r requirements.txt
    

    If you have problems installing openslide, please refer to the link: https://openslide.org/api/python/.

Usage

Even you have no idea about which patches belong to which WSI, just put all the patches in one folder and all the possible WSIs in another folder.

Then run the following command in your terminal ... done!

python match.py dir_wsi dir_patch [-lv LEVEL] [-ow ORIGINAL_WIDTH] [-oh ORIGINAL_HEIGHT] [-ds DSAMPLE] [-t THRESHOLD] [-fr FINETUNE_RANGE] [-ft FINETUNE_THRESH] [-v] [-o OUTPUT]
  • positional arguments:

    • dir_wsi: directory of whole slide images
    • dir_patch: directory of patch images
  • optional arguments:

    • -h, --help: show this help message and exit.
    • -lv, --level: level number for cropping patches from WSI, defualt: 0.
    • -ow, --original_width: original width of cropped patches, defualt: 3432.
    • -oh, --original_height: original height of cropped patches, defualt: 1884.
    • -ds, --dsample: downsampling rate for matching, default: 8.
    • -t, --threshold: threshold for SIFT algorithm, range: 0~1, default: 0.4.
    • -fr, --finetune_range: the pixel range for validating matched patch, default: 60.
    • -ft, --finetune_thresh: threshold for fine-tuning, range: 0~1, default: 0.9.
    • -v, --verbose: display cropping and patch image.
    • -o, --output: path of output file, default: "output.json".

Output

The output file will be saved as a json file and the format will look like this:

{
    "XXXX1.jpg": {
        "WSI name": "aabbcc.svs",
        "start coord": [23311, 8140]
    },
    "XXXX2.jpg": {
        "WSI name": "aabbcc.svs",
        "start coord": [20989, 11200]
    },
    "XXXX3.jpg": {
        "WSI name": "ccbbaa.tiff",
        "start coord": [24820, 9847]
    },
    .
    .
    .
}

"start coord" refers to the location (top left anchor) of patch in WSI of level 0.

API reference

You can also use the following API functions in an interactive environment like jupyter notebook.

match function

match.match(dir_wsi: str, dir_patch: str,
    level=0, ori_crop_size=(3432, 1884),
    dsample_match=8, threshold=0.4,
    finetune_range=60, finetune_thresh=0.9,
    debug_mode=False)
  • dir_wsi: string, directory of whole slide images.
  • dir_patch: string, directory of patch images.
  • level: integer, level number for cropping patches from WSI.
  • ori_crop_size: tuple of two integers, original size of cropped patches.
  • dsample_match: integer, downsampling rate for matching.
  • threshold: float (0~1), threshold for SIFT algorithm.
  • finetune_range: integer, the pixel range for validating matched patch.
  • finetune_thresh: float (0~1), threshold for fine-tuning.
  • debug_mode: boolean, whether to show cropping and patch image.

Returns

A dictionary contains matching information with the same format as output json.

save_json function

match.save_json(matched_dict: dict, path: str):
  • matched_dict: dictionary, the return from match().
  • path: string, path of output file.

Example

  1. Click the button below to download a little data from TCGA for testing:

"The patches and WSI here are based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga."

  1. Run the following command in terminal (replace <root> to yours):
python match.py <root>/PatchMatch/tcga_wsi  <root>/PatchMatch/tcga_patches -v

You are supposed to see some output like this:

Matching patches in slide: TCGA-55-6983-01Z-00-DX1.8f940a64-1f1b-4e6e-99ea-418175be2b3f.svs

thumbnail

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 351/351 [03:02<00:00,  1.93it/s]
Checking and fine-tuning patch: S005.jpg ...
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3600/3600 [00:03<00:00, 1150.13it/s]
min cross entropy: 348.647378096351
dist: 0.0
Matched!
Patch: S005.jpg is in slide: TCGA-55-6983-01Z-00-DX1.8f940a64-1f1b-4e6e-99ea-418175be2b3f.svs, start with: (23311, 8140).

matched_img

Checking and fine-tuning patch: S004.jpg ...
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3600/3600 [00:04<00:00, 781.99it/s]
min cross entropy: 325.29451424359223
dist: 0.0
Matched!
Patch: S004.jpg is in slide: TCGA-55-6983-01Z-00-DX1.8f940a64-1f1b-4e6e-99ea-418175be2b3f.svs, start with: (20989, 11200).
.
.
.
Matched 10/10 patches.
Saved output file to "output.json" successfully.

You can also click the button below to open the sample code in Colab:

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.