Miffka/lungs-finder

Library for detecting lungs on chest x-ray images for further processing. It is fast and able to work on embedded devices.

lungs-finder

Library for lungs detection on CXR images

Requirements:

• Python 3;
• OpenCV3.

Installation:

git clone git@github.com:Miffka/lungs-finder.git cd lungs-finder && pip3 install -e .

You can test your dataset using lungs_viewer.py.
It is a program that is displayed on the screenshot above.
Usage: lungs_viewer.py /your/dataset True
The last parameter defines whether labels will be displayed or not.

Usage

Function get_lungs

import cv2
import lungs_finder as lf

image = cv2.imread(YOUR_PNG_IMAGE_PATH, 0)
image = cv2.equalizeHist(image)
image = cv2.medianBlur(image, 3)
image = cv2.resize(image, (512, 512)) # resize to small scale is crucial, but 1024 works also okay

# Returns lungs (as is with padding)
found_lungs = lf.get_lungs(image, padding=20)

# Returns lung coordinates (x, y, width, height)
x, y, width, height = lf.get_lungs(image, padding=20, return_image=False)

# Tries to find square with lungs
found_lungs_square = lf.get_lungs(image, padding=20, square=True)

# Returns coordinates of square with lungs
x_sq, y_sq, width_sq, height_sq = lf.get_lungs(image, padding=20, square=True, return_image=False)

fig, ax = plt.subplots(1, 2, figsize=(10,10))

if found_lungs is not None:
    ax[0].imshow(found_lungs, cmap=mpl.cm.gray)
if found_lungs_square is not None:
    ax[1].imshow(found_lungs_square, cmap=mpl.cm.gray)

Visualize HOG, HAAR, LBP methods of finding lungs

from itertools import product
import os
import os.path as osp
import cv2

import matplotlib as mpl
import matplotlib.pyplot as plt
from matplotlib.patches import Rectangle

import lungs_finder as lf

def show_lungs(img_path, resize=512):
    image = cv2.imread(img_path, 0)
    image = cv2.equalizeHist(image)
    image = cv2.medianBlur(image, 3)
    image = cv2.resize(image, (resize, resize))
    found_lungs = lf.get_lungs(image, padding=20)
    found_lungs_square = lf.get_lungs(image, padding=20, square=True)

    fig, ax = plt.subplots(1,3, figsize=(15, 5))
    ax[0].imshow(image, cmap=mpl.cm.gray)
    ax[1].imshow(cv2.resize(found_lungs, (resize, resize)), cmap=mpl.cm.gray)
    ax[2].imshow(cv2.resize(found_lungs_square, (resize, resize)), cmap=mpl.cm.gray)


    for (side, method), color in zip(product(['right', 'left'],
                                             ['hog', 'lbp', 'haar']),
                                     ['yellow', 'red', 'blue']*2):
        out = eval(f'lf.find_{side}_lung_{method}')(image)
        if out is not None:
            x, y, width, height = out
            rect = Rectangle((x, y), width, height, fill=False, edgecolor=color)
            ax[0].add_patch(rect)

    fig.suptitle(osp.basename(img_path))
    plt.show();

for i, file in enumerate(os.listdir(osp.join(PATH_TO_YOUR_PNG_FILES))):
    show_lungs(osp.join(PATH_TO_YOUR_PNG_FILES, file), resize=512)
    if i > 50:
        break