An Introduction to Cerebral Infraction (Stroke) Prediction Model

Contents

  1. Introdution
  2. Preparing CSV
  3. Using End point
    • 3.1 Creating model
    • 3.2 Creating Endpoint Config
    • 3.3 Creating Endpoint
    • 3.4 Invoking Endpoint
    • 3.5. Deleting the Endpoint - Optional
  4. Model Output

1. Introdution

Stroke is a medical condition in which poor blood flow to the brain results in cell death. There are two main types of stroke: ischemic, due to lack of blood flow, and hemorrhagic, due to bleeding. Both result in parts of the brain not functioning properly.Prediction model helps to identify the person suffering from Stroke by taking into account various observational parameters and demographic details.

2. Preparing CSV

The csv file used in the model has 10 columns viz

  • AGE = Person's Age
  • 72095-3 = Activities of daily living score [HOOS]
  • 72093-8 = Quality of life score [HOOS]
  • 72016-9 = VR-12 Physical functioning (PF) score - oblique method
  • 72015-1 = VR-12 Role physical (RP) score - oblique method
  • 72014-4 = VR-12 Bodily pain (BP) score - oblique method
  • 72013-6 = VR-12 General health (GH) score - oblique method
  • 72012-8 = VR-12 Vitality (VT) score - oblique method
  • 72011-0 = VR-12 Social functioning (SF) score - oblique method
  • 72010-2 = VR-12 Role emotion (RE) score - oblique method

Same kind of CSV file should be fed to the model for best output.

For reference, A sample CSV file is attached in the link. The user only needs to clik on the link to get it downloaded to the local. The data should be in a similar format like the sample csv.

Download sample Data from the link : Data Source for Stroke

In [ ]:
import pandas as pd
df = pd.read_csv(' S3/Sagemaker path of Stroke.csv file')
df.to_csv("sample.csv",index=False)
df.head(3)

3. Using Endpoint

The model can be directly accessed via the console provided by AWS. However, for a more customized process one can access the model using the below code as well.

3.1 Creating Model

To create a model, import boto3, sagemaker and get the arn of the model package

In [ ]:
import boto3
import sagemaker
role = sagemaker.get_execution_role()
smmp = boto3.client('sagemakermp')
modelName='Name of the model'
modelArn = 'Model ARN name'
createHeatIndexResponse = smmp.create_model(ModelName=modelName,\
                             Containers=[{'ModelPackageName': modelArn}],\
                             ExecutionRoleArn=role,\
                             EnableNetworkIsolation=True )

3.2 Creating Endpoint Config

In [ ]:
configName ='<Input Configuration Name>'
instanceType = '<Input Instance Type>'
createHeatIndexEndpointConfig = smmp.create_endpoint_config(EndpointConfigName = configName, ProductionVariants = [{'InstanceType':instanceType, 'InitialInstanceCount':1, 'ModelName':modelName, 'VariantName':'xyz'}])

3.3 Creating Endpoint

In [ ]:
endpointName = '<Input Endpoint Name>'
createHeatIndexEndpoint = smmp.create_endpoint(EndpointName = endpointName, EndpointConfigName = configName)

3.4 Invoking Endpoint

In [ ]:
runtime = boto3.Session().client('runtime.sagemaker')

#Reading Input Data 
with open('sample.csv','rb') as f:
    payload = f.read()

response = runtime.invoke_endpoint(EndpointName = endpointName, ContentType = 'text/csv', Body = payload)
result = response['Body'].read().decode()

#Writing Output Data 
with open('sampleOutput.txt','w') as f:
    f.write(result)

3.5. Deleting the Endpoint - Optional

If you're ready to be done with this notebook, please run the delete_endpoint line in the cell below. This will remove the hosted endpoint you created and avoid any charges from a stray instance being left on

In [ ]:
sagemaker.Session().delete_endpoint(endpointName)

4. Model Output

In [ ]:
with open("./sampleOutput.txt","r") as f:
    sampleResponse =f.read()
    sampleResponse = sampleResponse.split('\n')
sampleResponse