In early August, 2024 several medical news services trumpeted the news that Artificial Intelligence can detect prostate cancer (PCa) on MRI with the same accuracy as radiologists:

• “AI model matches radiologists at detecting prostate cancer on MRIs” (Radiology Business, Aug. 7)
• “AI model as effective as radiologists in detecting prostate cancer on MRI” (MedPage Today, Aug. 6)
• “AI model matches radiologists in detecting clinically significant prostate cancer on MRI” (News Medical, Aug. 6)
• “AI model developed to detect prostate cancer found to perform like ‘experienced radiologist’” (Deccan Herald, Aug. 7)
• “AI model comparable to radiologists for finding prostate cancer on MRI” (AuntMinnie, Aug. 6)

The news broke within hours after the online publication of a Mayo Clinic study titled “Fully Automated Deep Learning Model to Detect Clinically Significant Prostate Cancer at MRI” (Cai et al., Aug. 6 2024).^[i]

According to Radiology published paper, the research team developed a Deep Learning (DL) model that was trained to predict the presence or absence of clinically significant PCa on multiparametric MRI (mpMRI) scans. The training and testing sets were derived from an initial pool of 7854 mpMRI scans for suspicion of PCa. After excluding cases that did not meet study criteria, the team arrived at 1514 MRI cases that were then biopsy-proven to have clinically significant PCa (that is, Gleason 7 or above), and 4221 MRIs without clinically significant PCa either by biopsy or PI-RADS 1 or 2.

These 5035 MRIs were used as the DL model’s training set; then 300 were randomly chosen for the test set, and 400 for the validation set. Next, 204 images were randomly selected for external validation by four experienced radiologists, each of whom interpreted 21 cases. Their performance at correctly predicting clinically significant PCa was compared with that of the DL model.

The results were virtually identical, with the DL model performing at the same level of accuracy as that of the radiologists. In fact, when the model’s predictions were paired with that of radiologists, the performance was even better than radiologists alone. Despite the headlines’ rather sensationalist tone, however, the authors of the study that ignited the press are not declaring—or even implying—that AI can now stand in for radiologists. They write, “We believe that our model has the potential to assist radiologists in identifying csPCa and facilitate lesion biopsy, hence improving the diagnosis of prostate cancer.”^[ii] Not only can DL increase PCa detection rates while reducing false positives, it can increase the speed and efficiency of radiologists’ work. Altogether, this can not only improve the quality of patient care and outcomes, it can also lower healthcare costs. The headlines may have gained attention by startling readers into mistakenly thinking AI might replace radiologists, but there is no mistake that as an adjunct to human radiologists, AI is good news for us all.

NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.

References

[i] Cai JC, Nakai H, Kuanar S, Froemming AT et al. Fully Automated Deep Learning Model to Detect Clinically Significant Prostate Cancer at MRI. Radiology. 2024 Aug;312(2):e232635.
[ii] Ibid.