Bland-Altman Plots(一致性评价)在python中的实现

假设有reader1 和reader2，分别对一定数量病人的某一影像指标进行评分，现在想看一下这两位研究者评分的一致性，绘制Bland-Altman图是一种较为直观、简单的方式。python代码实现方法如下：

首先读入数据

folderPath = "/Users/.../ICC/features4ICC/"

data1 = pd.read_excel(os.path.join(folderPath,"reader1_ap.xlsx"))
data2 = pd.read_excel(os.path.join(folderPath,"reader2_ap.xlsx"))

data1.insert(0,"reader",np.ones(data1.shape[0]))
data2.insert(0,"reader",np.ones(data2.shape[0])*2)

data1.insert(0,"target",range(data1.shape[0]))
data2.insert(0,"target",range(data2.shape[0]))

data = pd.concat([data1,data2])
print(data)

input_data.png

Method 1

import numpy as np
import pingouin as pg

ax = pg.plot_blandaltman(data1['original_shape_Elongation'], data2['original_shape_Elongation'])

method_1.png

Method 2

# pip install pyCompare # for the first time
import pyCompare

pyCompare.blandAltman(data1['original_shape_Elongation'], data2['original_shape_Elongation'], 
                      percentage=False, 
                      title='Bland-Altman Plot 2',
                      limitOfAgreement=1.96)

method_2.png

pyCompare.blandAltman(data1['original_shape_Elongation'], data2['original_shape_Elongation'],
            savePath='SavedFigure.tiff',
            figureFormat='tiff')

Method 3

import matplotlib.pyplot as plt
import numpy as np

def bland_altman_plot(data1, data2, *args, **kwargs):
    data1     = np.asarray(data1)
    data2     = np.asarray(data2)
    mean      = np.mean([data1, data2], axis=0)
    diff      = data1 - data2                   # Difference between data1 and data2
    md        = np.mean(diff)                   # Mean of the difference
    sd        = np.std(diff, axis=0)            # Standard deviation of the difference

    plt.scatter(mean, diff, *args, **kwargs)
    plt.axhline(md,           color='gray', linestyle='-')
    plt.axhline(md + 1.96*sd, color='gray', linestyle='--')
    plt.axhline(md - 1.96*sd, color='gray', linestyle='--')

from numpy.random import random

bland_altman_plot(data1['original_shape_Elongation'], data2['original_shape_Elongation'])
plt.title('Bland-Altman Plot 3')
plt.show()

method_3.png

参考资料：
Pingouin official documentation
Why & How to use the Bland-Altman plot for A/B testing | Python + code
Bland-Altman plot with confidence interval boundary in python