Tuesday, May 16, 2023

Artificial Intelligence Potentially Twice as Good at Screening for Lung Cancer than Humans Alone


As AI makes greater and greater inroads into our society and world, more and more health specialities are turning to AI to aid in diagnosis.

Now please note that above the phrase "to aid in diagnosis" was used.  Machines DO NOT make diagnostic decisions... only human beings can do that.  The role of any device is to help the doctor gain more data and more information.  Then, that data and information is used by the doctor to make a decision on how to proceed.  The decision rests with the human being and the human alone.

One of the things that some people have asked me about it this concept of machines or programs diagnosing.  My best explanation is to say that in order to make a sound clinical decision a good doctor will gather as much data as possible... and AI is simply another way to do that.  An AI system my remind the doctor to "look here".  That is the way the current dental systems work.  They analyze x-rays and then mark them so that a doctor can see areas that *might* be of concern.  That is simply providing data that the doctor should consider.

The article I read recently about lung cancer was interesting because it will also be gleaning data from x-rays.  The article states "In a real-world setting, machine learning-based software significantly boosted the identification of lung nodules on chest X-rays."

What that tells us is that these areas of concern are detected better by AI, but once again it only points out to the doctor areas to be more closely evaluated.  Machine Learning systems are really good at detecting changes in density of x-rays.  In the RGB color model for computing, there are 256 shades of gray while the accepted range of the human eye is somewhere from 16-32 shades.  Basically this is why AI systems can be a great help for doctors.  The computer is better are detecting nuanced changes.  By pointing these areas out to the doctor, diagnosis gets better and potentially earlier.

The article appeared in the publication Radiology and can be found by following this link.  

No comments:

Post a Comment