Python Web Application - Hypertension Status Prediction

Introduction :

Hypertension  ̶  or elevated blood pressure  ̶  is a serious medical condition that significantly increases the risks of heart, brain, kidney and other diseases.

An estimated 1.28 billion adults aged 30-79 years worldwide have hypertension, most (two-thirds) living in low- and middle-income countries

An estimated 46% of adults with hypertension are unaware that they have the condition.

Less than half of adults (42%) with hypertension are diagnosed and treated.

Approximately 1 in 5 adults (21%) with hypertension have it under control.

Hypertension is a major cause of premature death worldwide.

One of the global targets for noncommunicable diseases is to reduce the prevalence of hypertension by 33% between 2010 and 2030.

Source : WHO

Impact :

Early detection of hypertension can be a very effective measure to prevent any coplications associated with non managed increasing blood pressure. It can not only reduce the complication as well as unwanted medical treatment but also will prevent the degrading health statistics .

Problem Statement :

To develop an anlgorithm to predict the blood pressure status.

To use the demographic and life style data to flag the subject who may get hypertension. 

Approach :

ML model was developed 

1. Data Understanding

2. Data Cleaning and imputation

3. Data Vizualization

4. Data Preperation

5. Model building

6. Model Evaluation

HTML page was designed to get the user input

HTML page and Python was integrated to to execute Model on the data entered.

Output was displayed in HTML page as either of the following: 

• High Chance of Hypertension 

• Low Chance of Hypertension

Solution :

Once flagged with high chance of hypertension, subject can go under non pharmacological management to prevent any further complication as well as medical treatment. 

Non pharmacological management includes

• Lifestyle modification

• Eating habit

• Managing risk factors

Links :

import pandas as pd, numpy as np, seaborn as sns, matplotlib.pyplot as plt

from sklearn.linear_model import LogisticRegression

import statsmodels.api as sm

from flask import Flask, render_template, request

from sklearn.metrics import confusion_matrix

from sklearn import metrics

import webbrowser

from threading import Timer

app = Flask(__name__)

@app.route("/")

@app.route("/home")

def home():

    return render_template("health_index.html")

@app.route("/result", methods= ["POST", 'GET'])

def result():

    Rawdf = pd.read_csv("https://97221248-9bb3-4fc9-9d37-d1ad34bc7154.usrfiles.com/ugd/972212_590fd873a65640c1843bb2e745ec52fc.csv", sep=";")

    Rawdf['AgeinYr'] = (Rawdf.age)//365

    Rawdf['BP_Status'] = [1 if i > 140 or j > 90 else 0 for i,j in zip(Rawdf.ap_hi, Rawdf.ap_lo)  ]

    Finaldf = Rawdf[['gender','height','weight','cholesterol','alco','active','cardio','AgeinYr', 'BP_Status']]

    Finaldf['gender'] = Finaldf.gender.map({1:0,2:1})  # 1 - Male 0 - Female

    X_train = Finaldf.drop('BP_Status', axis=1)

    y_train = Finaldf[['BP_Status']]

    lr1 = sm.GLM(y_train, X_train, family= sm.families.Binomial())

    lm1 = lr1.fit()

    y_train_predict = lm1.predict(X_train)

    Pred_DF = pd.DataFrame({'Actual':y_train.values, 'Predict' : y_train_predict})

    Matrix_df = pd.DataFrame(columns=['Cut_off', 'accuracy', 'Sensitivity', 'Specificity'])

    cutt_off = list(np.arange(0,1,0.01))

    for i in cutt_off:

        val = [1 if j > i else 0 for j in Pred_DF.Predict]

        Pred_DF[i] = val

    for i in cutt_off:

        cm = metrics.confusion_matrix(Pred_DF.Actual,Pred_DF[i])

        total_val = sum(sum(cm))

        TP = cm[1,1]

        TN = cm[0,0]

        FP = cm[0,1]

        FN = cm[1,0]

        accuracy = (TP+TN)/total_val

        Sensitivity = TP / (TP+FN)

        Specificity = TN / (TN+FP)  

        Matrix_df.loc[i] = [i, accuracy, Sensitivity, Specificity]

    Matrix_df['Dif'] = abs(Matrix_df.Sensitivity - Matrix_df.Specificity)

    Matrix_df = Matrix_df.sort_values(by = 'Dif', ascending=False)

    Optimum_cutoff = Matrix_df[Matrix_df.Dif == Matrix_df.Dif.min()]

    ct_off = Optimum_cutoff['Cut_off'].values[0]

    output = request.form.to_dict()

    X_test = pd.DataFrame({'gender':int(output['gender']), 'height':int(output['height']), 'weight':int(output['weight']),

                            'cholesterol':int(output['chol']), 'alco':int(output['alcohol']), 'active':int(output['active']),

                            'cardio':int(output['cardio']), 'AgeinYr':int(output['age'])}, index=[0])

    y_test_pred = lm1.predict(X_test)

    if y_test_pred.values > 0.218:

        prediction = 'Chances of Hypertension'

    else:

        prediction = 'Low Chance of Hypertension'

    return render_template("health_index.html", prediction = prediction)

def open_browser():

      webbrowser.open_new('http://127.0.0.1:5000/')

if __name__ == '__main__':

    Timer(1, open_browser).start()

    app.run(port=5000)