042719

From_xml_Mask_batch.py

#!/usr/bin/env python3


#If you need more information about how ElementTree package handle XML file, please follow the link:
#https://docs.python.org/3/library/xml.etree.elementtree.html


#import multiresolutionimageinterface as mir
import matplotlib.pyplot as plt
import cv2
import numpy as np
import openslide
from pandas import DataFrame
import os.path as osp



#reader = mir.MultiResolutionImageReader()
#mr_image = reader.open('/home/wli/Downloads/tumor_036.tif')
#Ximageorg, Yimageorg = mr_image.getDimensions()
#dims = mr_image.getLevelDimensions(4)
#Ximage = (Ximage+240//2)//240
#Ximage = 4000
#Yimage = (Yimage+240//2)//240
#Yimage = 2000
import xml.etree.ElementTree as et
import pandas as pd
import glob
import math

def convert_xml_df(file):
    parseXML = et.parse(file)
    root = parseXML.getroot()
    dfcols = ['Name', 'Order', 'X', 'Y']
    df_xml = pd.DataFrame(columns=dfcols)
    for child in root.iter('Annotation'):
        for coordinate in child.iter('Coordinate'):
            Name = child.attrib.get('Name')
            Order = coordinate.attrib.get('Order')
            X_coord = float(coordinate.attrib.get('X'))
           # X_coord = X_coord - 30000
            #X_coord = ((X_coord)*dims[0])/Ximageorg
            X_coord = X_coord
            #X_coord = X_coord/32

            Y_coord = float(coordinate.attrib.get('Y'))
           # Y_coord = Y_coord - 155000
            #Y_coord = ((Y_coord)*dims[1])/Yimageorg
            Y_coord = Y_coord
            #Y_coord = Y_coord/32

            df_xml = df_xml.append(pd.Series([Name, Order, X_coord, Y_coord], index = dfcols), ignore_index=True)  # type: DataFrame
            df_xml = pd.DataFrame(df_xml)

    return (df_xml)



#x_values = list(annotations['X'].get_values())
#y_values = list(annotations['Y'].get_values())
#xy = list(zip(x_values,y_values))
def Remove_dup(duplicate):
    final_list = []
    for num in duplicate:
        if num not in final_list:
            final_list.append(num)
    return final_list



def mask_gen(final_list):

    coxy = [[] for x in range(len(final_list))]

    for index, n in enumerate(final_list):
        newx = annotations[annotations['Name']== n]['X']
        newy = annotations[annotations['Name']== n]['Y']
        print(n)
        print(newx, newy)
        newxy = list(zip(newx, newy))
        coxy[index] = np.array(newxy, dtype=np.int32)



    #image = cv2.imread('/home/wli/Downloads/tumor_036.xml', -1)
    canvas = np.zeros((int(dims[1]), int(dims[0])), np.uint8)
    #canvas = np.zeros((int(dims[1]/32), int(dims[0]/32)), np.uint8)

    #tile =mr_image.getUCharPatch(0, 0, dims[0], dims[1], 4)
    #canvas = np.zeros((Ximage, Yimage, 3), np.uint8) # fix the division
    #coords = np.array([xy], dtype=np.int32)

    #cv2.drawContours(canvas, [coords],-1, (0,255,0), -1)

    #cv2.drawContours(canvas, coxy, -1, (255, 255, 255), 10)
    #cv2.drawContours(canvas, coxy, -1, (255, 255, 255), CV_FILLED)
    cv2.fillPoly(canvas, pts = coxy, color=(255,255,255))
    #cv2.polylines(canvas, coxy, isClosed=True, color=(255,255,255), thickness=5)

    cv2.imwrite('/home/wzli/home_made_mask_files/%s.tif' % osp.basename(osp.splitext(xml_file)[0]), canvas)
#cv2.imshow("tile", tile);cv2.waitKey();cv2.destroyAllWindows()

#cv2.fillConvexPoly(mask, coords,1)
#mask = mask.astype(np.bool)

#output = np.zeros_like(image)
#output[mask] = image[mask]

#cv2.imshow('image',output)
#cv2.waitKey(0)
#cv2.destroyAllWindows()
if __name__ == '__main__':

    xml_folder  = '/home/wzli/Downloads/CAMELYON16/testing/lesion_annotations/'
    xml_paths = glob.glob(osp.join(xml_folder, '*.xml'))
    slide_folder = '/home/wzli/Downloads/CAMELYON16/testing/images/'
    #slide_paths = glob.glob(osp.join(slide_folder, '*.tif'))

    for xml_file in xml_paths:
        slide_name = osp.basename(xml_file.replace('.xml', '.tif'))
        slide_path = osp.join(slide_folder, slide_name)
        wsi_image = openslide.open_slide(slide_path)
        dims = wsi_image.dimensions
        annotations = convert_xml_df(xml_file)
        final_list = Remove_dup(annotations['Name'])
        mask_gen(final_list)