Artificial Intelligence in Medicine: Ultra-Fast MRI Accurately Detects Prostate Cancer

Today’s multiparametric MRI (mpMRI) has revolutionized the prostate cancer (PCa) world. Thanks to its high resolution, 3-dimensional imaging done by a powerful 3 Tesla (3T) magnet, experienced radiologists are able to identify areas that are suspicious for clinically significant PCa (csPCa). This helps avoid unnecessary biopsies, but if a biopsy is indicated, the area can be targeted with fewer needles.

Experts agree there is no downside to 3T mpMRI. There is, however, a slight drawback: 3T mpMRI scans take upwards of 30 minutes because they require three scanning sequences, or parameters:

• T2 weighted MRI shows anatomy (do not confused T2 weighted with 3 Tesla/3T)
• Diffusion weighted MRI shows the difference between normal tissue and cancer tissue
• Dynamic contrast enhanced MRI uses an intravenous contrast agent to detect unusual tumor blood flow.

The first two sequences are short, usually under 5 minutes each. They are considered the core sequences or “work horse” of prostate MRI. However, the sequence using the contrast agent requires more time to administer the agent and wait until it reaches the prostate gland before scanning. Thus, when all three sequences are used, scan time can take around 30 minutes—a situation that limits the number of patients who have access to prostate MRI during a busy radiology center day. Because of this, faster prostate scans are under study for new ways to save time without sacrificing accurate detection.

A common MRI method being tested is called biparametric MRI (bpMRI). Note: “bi” in front of word means two, while “multi” means more than two. Although the use of a contrast agent adds unique information that supplements the other two sequences, there are times when it is advisable or desirable to skip it. Thus, bpMRI uses only T2 weighted and diffusion weighted MRI, skipping the use of the third sequence using contrast. In turn, without the tumor blood flow information, it’s crucial that the image fidelity of bpMRI scans be at peak performance so radiologists get the most accurate read possible.

How artificial intelligence can help

In Jan. 2025, a team of Swiss researchers tested a new bpMRI method that incorporates artificial intelligence (AI) to enhance imaging features. For their study they incorporated a machine learning program that could “significantly reduce image noise, enhance signal-to-noise ratio, and optimize image sharpness, consequently reducing acquisition time without compromising image quality. ”^[i] In fact, previous studies have shown that AI brings lightning light speed to enhancing bpMRI image features, reducing total scan time to 5 minutes or less. This is called ultra-fast MRI.

With machine learning now integrated into processing their scanning, the team tested its performance with 123 patients who had not had a previous biopsy. Upon enrollment, all patients had a conventional mpMRI protocol as well as an ultra-fast bpMRI protocol. Any csPCa detected on scans was then biopsied in order to obtain actual diagnosis results. Then 2 radiologists who were blinded to the biopsy results read all mpMRI and bpMRI scan results without knowing which patient individual scans came from. One radiologist had 4 years’ experience, the other had 3 years’ experience. The radiologists’ interpretations were compared with biopsy results, and with each other.

Results

Both scanning protocols had approximately the same diagnostic accuracy when imaging was compared with biopsy findings. The more experienced radiologist had slightly better accuracy than the less experienced one. The real difference was revealed when mpMRI was compared with bpMRI. The authors report, “The ultra-fast bpMRI protocol had significantly better image quality ratings … and achieved a reduction in scan time of 80% compared to conventional mpMRI.” Therefore, they concluded, “Deep-learning-assisted ultra-fast bpMRI protocols offer a promising alternative to conventional mpMRI for diagnosing csPCa in biopsy-naïve patients with comparable inter-reader agreement and diagnostic performance at superior image quality.”^[ii] At the same time, they point out that more experienced readers have better diagnostic performance.

The findings of this team are promising when the use of contrast is not available or desirable. Meanwhile, European research increasingly points to the value of joining PSA plus mpMRI for screening purposes in a large population. Their results show that while this may sound more expensive in the short fun, in the long run it saves healthcare dollars by greatly reducing the number of men who are over biopsied, which leads to overtreatment for many of them, which leads to an economic burden of both treatment costs and managing years of after-effects. If ultra-fast bpMRI were the method of choice for men whose PSA is abnormal, the availability of imaging to clarify a high PSA in hundreds or thousands of men becomes much more feasible if scan times are greatly reduced. All this, thanks to AI integrated into prostate MRI.

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] Pausch AM, Filleböck V, Elsner C, Rupp NJ et al. Ultra-fast biparametric MRI in prostate cancer assessment: Diagnostic performance and image quality compared to conventional multiparametric MRI. Eur J Radiol Open.
2025 Jan 21;14:100635.
[ii] Ibid.