If you’ve ever been involved in a simulation game, or a simulation training model, you are interacting with something that’s not really there but feels like it is. In fact, during the experience the same neurons in the brain are activated as if the simulation IS the real world. Thus, the brain can be functionally educated to do something, without actually doing it in the physical world. For instance, aspiring pilots can acquire many of the flight skills they need by training on a flight simulator before they take the controls of a real plane. They’ve gained practice with fewer risks while also saving time and money.

In January, 2025 the journal Radiology published a study by Huang, et al. entitled “Deep Learning to Simulate Contrast-Enhanced MRI for Evaluating Suspected Prostate Cancer.”^[i] The authors created and tested a Deep Learning (DL) program to simulate an MRI scanning sequence (parameter) called Dynamic Contrast Enhanced (DCE) MRI. It is used to identify tumor blood flow in an area of the prostate gland suspicious for clinically significant prostate cancer (csPCa).

DCE MRI is used with two other sequences, T2 weighted (T2 MRI) and Diffusion Weighted (DW MRI or DWI); thus, multiparametric MRI (mpMRI) means the combining of T2, DWI and DCE MRI to gain complementary information about clinically significant prostate cancer (csPCa). However, it requires an added intravenous contrast agent. This adds more cost and time to the scan; plus, if a patient has poor kidney function he can’t have the agent which is formulated with a heavy metal called gadolinium. A weak or compromised kidney may not be able to cleanse the agent from the body.

Given the added time, expense, and possible risk to certain patients, there have been studies testing biparametric MRI, which uses only two parameters (T2 and DWI), not DCE. While this reduces the duration of the scan as well as the cost, DCE is a “value added” sequence because it reveals information (tumor blood flow) that the other two parameters don’t pick up, and helps evaluate disease aggression.

Therefore, the Huang team created and tested a computer program that simulates a DCE sequence without actually administering a contrast agent. According to their paper, they had complete multiparametric MRI scans from 567 patients suspected of having prostate cancer. They used 244 complete mpMRI scans to train a Deep Learning model. They then tested the model on the remaining 323 patient scans with the original acquired DCE images eliminated. Based on biparametric results, could the simulation model predict the DCE results without “seeing” the original live contrast?

To find out, the research team had three experienced radiologist readers each interpret the 323 test simulation scan results (“synthetic” mpMRI because the DCE component was simulated) by assigning a PI-RADS score to each. The PI-RADS scale from 1-5 predicts the presence of csPCA; the higher the number, the greater the probability. The authors write, “Simulated and acquired contrast-enhanced images demonstrated high similarity … with excellent reader agreement of PI-RADS scores. … When simulated contrast-enhanced imaging was added to biparametric MRI, 34 of 323 (10.5%) patients were upgraded to PI-RADS 4 from PI-RADS 3.”^[ii] In other words, the experiment demonstrated success!

According to an editorial in the same journal, the Huang team found “…excellent agreement between the synthetic data and the acquired ground-truth dynamic contrast-enhanced images, thereby raising the promise of generative AI providing enhancement information without the administration of gadolinium-based contrast material.”^[iii]

Therefore, the Deep Learning program offers the potential to supplement biparametric MRI with the important identifying tumor information provided by DCE—without the extra time and cost of mpMRI analysis. Just as important, it avoids risks for particular patients who can’t tolerate the contrast agent. This is one more way in which AI benefits prostate cancer patients. We invite you to explore more blogs in our series on Artificial Intelligence and other topics.

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] Huang H, Mo J, Ding Z, Peng X, Liu R et al. Deep Learning to Simulate Contrast-Enhanced MRI for Evaluating Suspected Prostate Cancer. Radiology. 2025 Jan;314(1):e240238.
[ii] Ibid.
[iii] Neji R, Goh V. Toward Replacing Contrast Agents in Prostate MRI Using Generative Artificial Intelligence. Radiology. 2025 Jan;314(1):e243287.