Machine Learning Helps to Grok Blood Test Results


Historically getting a blood test has been a rather cumbersome process. Typically it required visiting a clinic, receiving a prescription from a doctor (or equivalent), going to another clinic for the blood draw and finally receiving the results. The results were offered in a printed form with short acronyms and reference ranges. Little information was provided to help navigate the meaning of the results. Finally, the doctor would review the results with the patient during a follow up visit, but the interpretation is limited the doctors wisdom. This process is expensive, time consuming and provides very little insight as to what ailments the patient could be dealing with.

The emerging process for getting a blood test which has been quickly embraced by biohackers all over the planet is a total 180 degree flip from the old way. Go to a website such as BloodCalculator, order a $115 blood chemistry test+report and visit a lab near your for the blood draw. This test will be available within a week in the form of a full online report that offers detailed analysis based on the blood chemistry results and predictive analysis of things that have not been directly tested for. The standard medical process for blood work is now optional.

BloodCalculator uses a GPU powered machine learning engine (XGBoost). It can a provide a patient with direction for what to look into. The reports are complete with optimal ranges (versus just normal), explanations of how different values relate to each other and neatly categorized by topics such as (immune, vitamins, toxic metals, hormones, nutrients, etc.). The machine learning engine was trained using data obtained from 76,000 patients. Looking at a sample test the engine was 100% confident in it’s finding that a patient had pinworms and hepatitis-B.

The most exciting aspects of blood testing through this mostly online process are the following:

  • knowledge – the patients owns their data, learns from reports and has time to research the findings in detail
  • convenience – going to one website and making a single trip to a laboratory greatly speeds up the process of getting tested
  • savings – avoiding health insurance billing issues, getting predicted values that were not tested and forecasts for conditions that would be unusual to test for can save a lot of money

 

 


via Adafruit

Thoughts or comments?