Luna16数据集转VOC数据集&肺实质分割&生成,Mat.zip

import SimpleITK as sitk import matplotlib.pyplot as plt import numpy as np import torch.nn as nn import torch from PIL import Image import matplotlib.image as mpimg # mpimg 用于读取图片 import imageio from skimage import morphology from skimage import measure from sklearn.cluster import KMeans from skimage.transform import resize """ import scipy.misc scipy.misc.toimage(image_array).save('outfile.jpg') """ import cv2 """ x= (x_ano-x_offset)/x_ElementSpacing y= (y_ano-y_offset)/y_ElementSpacing z= (z_ano-z_offset)/z_ElementSpacing 其中，x是实际对应三维矩阵中的坐标。 x_ano是肺结节在annotation.csv中的坐标. x_offset是质心坐标. x_ElementSpacing是在x轴方向上的步长。相当于每一个像素对应现实世界中的长度。 """ #定义CT图像类保存 中心点 class CTImage(object): """docstring for Hotel""" def __init__(self, x_offset, y_offset, z_offset, x_ElementSpacing, y_ElementSpacing, z_ElementSpacing): self.x_offset = x_offset self.y_offset = y_offset self.z_offset = z_offset self.x_ElementSpacing = x_ElementSpacing self.y_ElementSpacing = y_ElementSpacing self.z_ElementSpacing = z_ElementSpacing self.ElementSpacing = np.array([x_ElementSpacing, y_ElementSpacing, z_ElementSpacing]) def load_itk_image(filename): with open(filename) as f: contents = f.readlines() line = [k for k in contents if k.startswith('TransformMatrix')][0] offset = [k for k in contents if k.startswith('Offset')][0] EleSpacing = [k for k in contents if k.startswith('ElementSpacing')][0] # 把值进行分割 offArr = np.array(offset.split(' = ')[1].split(' ')).astype('float') eleArr = np.array(EleSpacing.split(' = ')[1].split(' ')).astype('float') CT = CTImage(offArr[0],offArr[1],offArr[2],eleArr[0],eleArr[1],eleArr[2]) transform = np.array(line.split(' = ')[1].split(' ')).astype('float') transform = np.round(transform)#round() 方法返回浮点数x的四舍五入值 if np.any(transform != np.array([1, 0, 0, 0, 1, 0, 0, 0, 1])):#判断是否相等 isflip = True else: isflip = False itkimage = sitk.ReadImage(filename) numpyimage = sitk.GetArrayFromImage(itkimage) if(isflip == True): numpyimage = numpyimage[:,::-1,::-1] #::-1 倒序 return (numpyimage,CT,isflip) def truncate_hu(image_array): image_array[image_array > 400] = 400 image_array[image_array <-1000] = -1000 def normalazation(image_array): max = image_array.max() min = image_array.min() #归一化 image_array = (image_array - min)/(max - min)*255 #image_array = image_array.astype(int)#整型 image_array = np.round(image_array) return image_array #找到较大连通域区域的索引 def findMaxRegion(img): #计算连通域 img = img.astype(np.uint8) num, labels = cv2.connectedComponents(img, connectivity = 4) #找到连通域的最大值 getLabel = [0] for i in range(1, num): getLabel.append(np.sum(labels==i)) #求得最大连通域的索引 maxNum = 0 for i in getLabel: if i > maxNum: maxNum = i getIndex = [] for i in range(num): if getLabel[i] > maxNum/2: getIndex.append(i) maskLabelInedx = [np.where(labels == x) for x in getIndex ] #生成mask maskLabel = np.zeros(labels.shape, dtype = int) for i in range(len(maskLabelInedx)): maskLabel[maskLabelInedx[i]] = 1 return maskLabel def fill_color_demo(image): copyImage = image.copy() h, w = copyImage.shape[:2] mask = np.zeros([h+2, w+2], np.uint8) #这里必须为 h+2,w+2 cv2.floodFill(copyImage, mask, (20, 20), (0, 125, 125), (100, 100, 100), (50, 50, 50), cv2.FLOODFILL_FIXED_RANGE) cv2.imshow("fill_color_demo", copyImage) def getLungMask(imNoudle): img = imNoudle #1. 标准化数据 #Standardize the pixel values mean = np.mean(imNoudle) std = np.std(imNoudle) imNoudle = imNoudle-mean imNoudle = imNoudle/std #查看数值分布 f, (ax1, ax2) = plt.subplots(1, 2,figsize=(8,8)) ax1.imshow(imNoudle,cmap='gray') plt.hist(imNoudle.flatten(),bins=200) plt.show() #cv2.imshow("1",imNoudle) #找出肺部附近的平均像素值，对洗掉的图像进行重新正规化 middle = imNoudle[100:400,100:400] mean = np.mean(middle) #使用Kmeans分离前景(不透明组织)和背景(透明组织，即肺) #这样做只在图像的中心，尽量避免非组织部分的图像 #np.prod()函数用来计算所有元素的乘积 kmeans = KMeans(n_clusters=2).fit(np.reshape(middle,[np.prod(middle.shape),1])) centers = sorted(kmeans.cluster_centers_.flatten()) threshold = np.mean(centers) thresh_img = np.where(imNoudle<threshold,1.0,0.0) # threshold the image #cv2.imshow("thresh_img",thresh_img) image_array = thresh_img plt.imshow(image_array,cmap='gray') plt.show() #最初的侵蚀对去除某些区域的颗粒状很有帮助，然后使用大的膨胀使肺区域吞噬血管 erosion = cv2.erode(thresh_img, np.ones((4, 4), np.uint8)) # 腐蚀 dilation = cv2.dilate(erosion, np.ones((10, 10), np.uint8)) # 膨胀 #dilation = dilation * np.ones(dilation.shape, dtype = np.uint8) #cv2.imshow("after change",dilation) #eroded = morphology.erosion(thresh_img,np.ones([4,4])) #dilation = morphology.dilation(eroded,np.ones([10,10])) #在skimage包中，使用measure子模块下的label函数即可实现连通区域标记。 #参数input表示需要处理的二值图像，connectivity表示判定连通的模式（1代表4连通，2代表8连通）， #输出labels为一个从0开始的标记数组。 labels = measure.label(dilation) fig,ax = plt.subplots(2,2,figsize=[8,8]) ax[0,0].imshow(thresh_img,cmap='gray') ax[0,1].imshow(erosion,cmap='gray') ax[1,0].imshow(dilation,cmap='gray') ax[1,1].imshow(labels) # 标注mask区域切片图 plt.show() #np.unique 该函数是去除数组中的重复数字，并进行排序之后输出 label_vals = np.unique(labels) #measure.regionprops筛选连通区域 regions = measure.regionprops(labels) good_labels = [] """ area int 区域内像素点总数 bbox tuple 边界外接框(min_row, min_col, max_row, max_col) centroid array　　 质心坐标 convex_area int 凸包内像素点总数 convex_image ndarray 和边界外接框同大小的凸包　　 coords ndarray 区域内像素点坐标 Eccentricity  float 离心率 label int 区域标记 """ for prop in regions: B = prop.bbox if B[2]-B[0]<475 and B[3]-B[1]<475 and B[0]>40 and B[2]<472: good_labels.append(prop.label) mask = np.ndarray([512,512],dtype=np.int8) mask[:] = 0 #这里的mask是肺用的，不是肺结节用的，在只剩下肺之后， #我们再做一次大的膨胀来填充和取出肺部mask for N in good_labels: mask = mask + np.where(labels==N,1,0) mask = mask.astype(np.uint8) #mask = cv2.erode(mask, np.ones((10, 10)))#腐蚀会造成一些空洞 mask = cv2.dilate(mask, np.ones((10, 10)))#膨胀填充空洞 #mask = morphology.erosion(mask,np.ones([10,10]))#腐蚀 滤除一些边缘 maskLabel = findMaxRegion(mask) fig,ax = plt.subplots(2,2,figsize=[10,10]) ax[0,0].imshow(img) # CT切片图 ax[0,1].imshow(img,cmap='gray') # CT切片灰度图 ax[1,0].imshow(maskLabel,cmap='gray') # 标注mask，标注区域为1�