Section 4-6 of the first week of the second lesson: Appreciation of medical prognosis cases + homework analysis

second lesson first week4-7节 Appreciation of medical prognosis cases+作业解析

视频地址：B > Tina-姐

预后的案例

在这节课中,We will look at prognostic clinical examples.看看预后任务的输入和输出是什么样子的,以及我们如何权衡输入之间的权重.

我们可以把预后模型看作一个系统,它把病人的资料（profile）作为输入,并输出病人的风险评分.

profile 可以包括：

• 临床病史(包括任何以前的治疗过程中的重大疾病)
• 体检结果,如体温和血压等生命体征.
• 实验室检查,如全血细胞计数和包括CT扫描在内的成像等.

预后模型可以取其中的一个,并为患者输出一个风险评分.现在,风险评分可能是任意数字,也可能是概率.

让我们看一个预后模型的例子.这个预后模型将使用两个患者特征来得出心脏病的风险评分.

假设我们有一个75岁以上的吸烟者,我们会给吸烟一个分数,超过75岁的人特征的一个分数,最后一个风险评分是2.

让我们让这个预测模型更有趣一点.假设75岁以上的人比吸烟的风险高出一倍,我们希望模型能给年龄赋予两倍于吸烟的权重.

然后我们可以指定一个权重,或者如果你是吸烟者特征1和年龄特征2的系数.现在,我们可以用这些系数乘以这个值,得到每个特征的贡献,1代表吸烟,2代表75岁以上,总风险评分为3.

接下来我们看3个案例

• Risk calculator for patients with atrial fibrillation

Atrial fibrillation is a common cardiac arrhythmia,Can make patients at a stroke（stroke）的风险.For patients with atrial fibrillation,We'll look at an example of a model that predicts stroke risk within a year.这被称为CHA2DS2-VASc评分,is an abbreviation for a characteristic used as part of a patient profile, The number represents the weight of that feature.

We will count a70aged male with atrial fibrillationCHA2DS2-VASc评分,The patient has hypertension and diabetes.

Our first task is to enter the value for that patient into the table.We know this patient has high blood pressure and diabetes,know his age

So the risk score is3

• liver disease mortality

In our second example,we will study mortality.We will study one for12Scores of patients on the liver transplant waiting list over the age of.This score can estimate the patient3month mortality,It is also one of the factors that determines the speed at which a patient receives a liver transplant..

The figure below is a model of end-stage liver disease.,产生所谓的MELD评分.

We will again of manual calculationMELD评分.在这里,我们有一个50岁的妇女,she has some lab results,如下图.

This mod has two new features to make it even more interesting.首先,请注意,The first characteristic is not only a laboratory values,but the value oflog对数.When there is reason to believe that the relationship between risk and characteristic is linear in the natural logarithm of the characteristic,Usually the natural logarithm of the feature is used instead of the feature itself.

The second thing to note here is that there is a cutoff value.The intercept corresponds tovalue总是1,So if the other eigenvalues ​​are0,We can think of the intercept as the expected risk score（最低为0.643）.

现在,Let's calculate this patient'sMELD评分.

请注意,得到的10score does not tell us directly3Probability of Survival in Months,But with other patientsMELD评分相比,it is informative.

• risk of heart disease
  在我们的最后一个例子中,我们将研究20岁或20Patients over the age of 2 who do not yet have heart disease10Risk of heart disease during the year.

这是ASCVDRisk assessment is.We will again manually calculate the patient'sASCVD评分.

首先,请注意,In addition to taking some characteristics of natural logarithms.We have a product of two characteristics,For example, the natural logarithm of age multiplied byHDL-C的自然对数.这些被称为交互项,We'll dive into why they're useful.

But what needs to be understood now is that,For interactive items,We want to multiply two values.在这种情况下,We will take the natural logarithm of age and multiply byHDL-C的自然对数.这里要注意的第二件事是,we have a negative coefficient associated with some features.例如,we have aHDL-CThe negative coefficient of the natural log correlation of,A negative coefficient here means that the contribution of this feature is negative,This means that it reduces the score.这是有道理的,因为HDL-Cis high-density cholesterol,commonly known as good cholesterol.因此,We can expect to reduce the risk of heart disease.

We can calculate the risk score as in the previous example.在这里,We use the information we have to fill in the values.Then multiply each value by the coefficient,and add this column,get the sum of the actual values.


Convert this sum to a risk score using the formula below.

Don't worry too much about where this formula comes from.All that is important to understand is that,We can substitute this formula.draw this patient10The probability of developing heart disease during the year is3%.

Assignments are set up for these three cases after this section

作业解析

在这里,学习使用 Python Function Implementation Risk Score.包括

• 心房颤动：Chads-vasc 评分
• 肝病：MELD评分
• 心脏病：ASCVD 评分

作业文件名：C2_W1_lecture_ex_02.ipynb
作业地址：gongzhonghao > 菜单栏 > Find it in Wu Enda

This work is done by buildingpythonThe function implements the score calculation on the above picture.But features and feature weights are artificially defined.does not match the actual application,因此,仅做参考,作业不难.5Solved in minutes.

如 Chads-vasc 评分:Customize a scoring function

def chads_vasc_score(input_c, input_h, input_a2, input_d, input_s2, input_v, input_a, input_sc):
    # congestive heart failure  充血性心力衰竭
    coef_c = 1 
    
    # Coefficient for hypertension Hypertension coefficient
    coef_h = 1 
    
    # Coefficient for Age >= 75 years
    coef_a2 = 2
    
    # Coefficient for diabetes mellitus  Diabetes factor
    coef_d = 1 
    
    # Coefficient for stroke  中风
    coef_s2 = 2
    
    # Coefficient for vascular disease   Vascular Disease Coefficient
    coef_v = 1
    
    # Coefficient for age 65 to 74 years     
    coef_a = 1
    
    # TODO Coefficient for female  Coefficient of women
    coef_sc = 1
    
    # Calculate the risk score
    risk_score = (input_c * coef_c) +\
                 (input_h * coef_h) +\
                 (input_a2 * coef_a2) +\
                 (input_d * coef_d) +\
                 (input_s2 * coef_s2) +\
                 (input_v * coef_v) +\
                 (input_a * coef_a) +\
                 (input_sc * coef_sc)
    
    return risk_score

Call the function to get the score

tmp_c = 0
tmp_h = 1
tmp_a2 = 0
tmp_d = 0
tmp_s2 = 0
tmp_v = 1
tmp_a = 0
tmp_sc = 1

print(f"The chads-vasc score for this patient is",
      f"{
      chads_vasc_score(tmp_c, tmp_h, tmp_a2, tmp_d, tmp_s2, tmp_v, tmp_a, tmp_sc)}")

文章持续更新,可以关注微信公众号【医学图像人工智能实战营】获取最新动态,一个关注于医学图像处理领域前沿科技的公众号.坚持已实践为主,手把手带你做项目,打比赛,写论文.凡原创文章皆提供理论讲解,实验代码,实验数据.只有实践才能成长的更快,关注我们,一起学习进步~

我是Tina, 我们下篇博客见~

白天工作晚上写文,呕心沥血

觉得写的不错的话最后,求点赞,评论,收藏.或者一键三连