Healthcare Short: Alzheimer’s Predictive Model and Intervention

Posted by

This post has been republished via RSS; it originally appeared at: Healthcare and Life Sciences Blog articles.

Story Background

​​​​​​​Alzheimer’s disease is one of the top 10 leading causes of death in the United States. The 5th leading cause of death among adults aged 65 years or older. In 2020, an estimated 5.8 million Americans aged 65 years or older had Alzheimer’s disease. In 2010, the costs of treating Alzheimer’s disease were projected to fall between $159 and $215 billion.  The science of risk reduction is quickly evolving, and major breakthroughs are within reach. There is growing scientific evidence that healthy behaviors, which have been shown to prevent cancer, diabetes, and heart disease, may also reduce the risk for subjective cognitive decline.

 

 

Business Challenge 

Alzheimer's is a type of dementia that affects memory, thinking, and behavior. A term for memory loss and other cognitive abilities serious enough to interfere with daily life. Medical advancements over the years led to significant gains in life expectancy, however, one of the downsides of this feat is the increase in cases of dementia which is expected to continue into the future. Early detection of dementia can help prevent or delay complications and improve interventions to slow down the progression.

 

Business Outcomes 

Decrease in overall healthcare burden. Increase in daily living activity, the life span of members, and quality of life.

Enabled by:

  • Model accuracy is 93%.
  • The area Under Curve is 99%
  • F1 Score is 93%
  • Interventions to slow the progression of Alzheimer’s disease.

 

Solution Overview 

This model was designed to predict various stages of Alzheimer’s using imaging data, then take necessary action depending on the output

  • Use representative images including mild, moderate, very mild, and non-dementia cases to train and test the model.
  • Predict the class as a target and select an algorithm with favorable performance metrics.
  • Deploy champion model in Azure machine learning workspace for production.
  • Run the model with net new images to determine the class and apply a specific variation of the use case. 

 

KEY FEATURES – Variations of Use Case

  • Parallel assistance and confirmatory test to Radiologists in the reading of images to enhance conclusions.
  • Postmortem analysis of images to verify what diagnosis might have been missed in retrospect and then outreaching those individuals.
  • Net new secondary diagnosis – opportunities aside from the primary reason for images.
  • The extensible solution to other diseases using various images.

 

Please follow the aka.ms/HLSBlog for all this great content. 

Thanks for reading, Shelly Avery | Email, LinkedIn  

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.