Online Program

Return to main conference page
Friday, May 31
Machine Learning
Machine Learning E-Posters, I
Fri, May 31, 9:45 AM - 10:45 AM
Grand Ballroom Foyer

Artificial Intelligence Mammography Model and Healthcare Savings Opportunity (305138)

*Olajide Israel Ajayi, Blue Cross NC 

Keywords: Machine Learning, Artificial Intelligence, Mammography, imaging


Artificial Intelligence Mammography Model and Healthcare Savings Opportunity Olajide Ajayi, Blue Cross NC

The traditional approach to imaging services which typically involves examination and confirmation of images to ensure accuracy is time consuming. Although the actual scanning takes few minutes, total length of procedure is around 1 hour and in some complex cases a lot longer and possible diagnostic errors.

A new model is hereby proposed with the intent to significantly reduce duration of procedure and get more accurate as well as sufficient quality of conclusion from images. The advantages of this new model cannot be over-emphasized; which includes - Image processing capability being better than any expert/trained eye, Radiologist/Technician, Robust pool of historical images compared to any one singular health system or provider, ability to learn faster from mistakes - less diagnostic errors (most common type of medical mistake according to U.S. Department of Health and Human Services/Agency for Healthcare Research and Quality - June, 2017). The initial deployment of this model includes a parallel run with human overrides due to limited success stories using this model.

The findings of this research using various models such as Multivariate Statistics - Discriminant Analysis and Artificial Neural Network Algorithm via SVM/Random Forest approach on blood work as well as imaging data yielded sensitivity/specificity/accuracy in the neighborhood of 99%. The result of final model with a run time of two minutes includes ability of the AI to read out result of mammography test to patients using text to speech package in R. Currently, a lot of research is on-going to make it communicate with humans and extend this study to all radiology codes as well as cardiovascular diseases.