A new prostate cancer detection tool that uses artificial intelligence to evaluate MRI scans has been found to be about as accurate as an experienced radiologist, new research reveals.
Giving Radiologists a Run for Their Money
According to a new study published in the journal IEEE Transactions on Medical Imaging, a new AI-based prostate cancer detection method known as FocalNet is surprisingly accurate when it comes to doing its job, raising high hopes that the new detection method will become a valuable tool in the future in diagnosing prostate cancer patients properly.
FocalNet, which was developed by a team of researchers from UCLA, turns out to be quite startlingly accurate in its ability to read and analyze MRI scan data to make diagnostic decisions. Its performance, in fact, rivals that of some human specialists. While a good radiologist can be expected to get things right around 83.9 percent of the time, FocalNet’s accuracy was tested at 80.5 percent. Not bad for a computer program!
How FocalNet Works
The core of how the new AI detection system works isn’t necessarily complex. FocalNet is a neural network that runs an algorithm that compares MRI data against a database to determine if the new data indicates a high likelihood of prostate cancer. What is complex, however, is how FocalNet’s algorithm works; there are more than a million trainable variables the program compares new data against.
UCLA researchers “trained” FocalNet by constructing a database of 417 MRI scans from men diagnosed with prostate cancer. These scans made it possible for the neural network to both assess tumors and classify them consistently, which is part of why FocalNet can compare new data and categorize it with such high accuracy, especially when combined with such a high number of variables used as comparison points.
Not Exactly Automatic Diagnosis
It’s true that FocalNet’s accuracy is astoundingly high, but if you’re afraid your radiologist is about to be replaced by a machine you don’t have to worry just yet. The new AI tool is meant to act as a supplement, not a replacement. As a tool to support less-experienced radiologists, FocalNet could be used to great effect to both provide diagnostic support as well as speed up the process significantly, resulting in higher speed and accuracy in the future. For an experienced radiologist, this tool is a highly accurate second layer of confirmation to reinforce their initial diagnosis.