Prostate cancer detection boosted with computer assistance
The addition of computer-aided diagnostic generated MRI series could help radiologists identify clinically significant prostate cancer more frequently.
This was the case in a study that was recently shared in Academic Radiology. In it, researchers detailed how radiologists who used a CAD generated MRI series had improved diagnostic performance in comparison to when they did not [1]. Authors of the new paper explained how use of such a tool can improve the diagnostic performances of the Prostate Imaging Reporting and Data System (PI-RADS) classification for prostate cancer detection, which are known to differ based on reader experience, care setting, MRI field strength, scan protocols, etc.
“This CAD device produces a color map overlaid on the axial T2W series to highlight regions within the prostate that are suspicious for cancer,” corresponding author Mark A. Anderson, MD, of the Department of Radiology, Abdominal Imaging Division, at Massachusetts General Hospital, and colleagues wrote, adding that the device “also produces a numeric value for each detected lesion that represents the likelihood of prostate cancer.”
The device’s use also does not require the use of contrast-enhanced sequences, which benefits patients who cannot receive contrast, the authors shared.
For the study, nine radiologists interpreted 150 prostate MRI exams (not included in the model’s development) with and without random forest-based CAD model-generated MRI series. Without CAD, inter-rater agreement for PI-RADS scores was moderate, but significantly improved to substantial with CAD. The experts reported that CAD also improved average AUC. The greatest improvements were seen in the transition zone, according to the data.
“Tools which can increase the reproducibility of PI-RADS scoring while improving accuracy, particularly in the transition zone, are highly desirable,” the authors wrote.
Additionally, the authors noted that the study subjects involved MRI exams from multiple institutions, scanner vendors, were acquired at 3T and 1.5T field strengths, with and without endorectal coils, and variable maximum b-value—all factors that they believe make their findings translatable to a wider demographic of both readers and patients.
To learn more about the study, click here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.