AI Improves Identification of Extracapsular Extension Before Surgery

We’ve posted numerous blogs on the contributions of Artificial Intelligence (AI) to various aspects of MRI detection and diagnosis of prostate cancer (PCa). We’ve reported AI’s ability to predict metastasis, assign Gleason scores and PI-RADS scores, pinpoint tumor margins for focal therapy, rule out need for biopsy, predict PCa metastasis, and more—showing the accuracy and feasibility of submitting MRI scans for AI analysis. The number of AI models and programs is rapidly growing, which keeps developers and researchers very busy testing them.

Now we’re happy to describe the latest published results of an AI study by a team out of UCLA. They tested a program called Unfold AI which they applied to the task of identifying PCa patients at risk of extracapsular extension (ECE), meaning that the tumor has broken through the outer edge (capsule) of the gland and extends into the soft tissue surrounding it.

The problem with ECE

PCa patients diagnosed with very low to low-risk disease that is not aggressive generally have the widest range of treatment options available to them, including Active Surveillance or focal therapy. However, for patients with intermediate to high risk PCa, radical prostatectomy (RP) is the most common recommendation. Such patients tend to have higher volume, more aggressive disease. This means there is a greater chance that it may have grown and extended beyond the capsule. ECE carries more risk of recurrence, and may mean a patient is not a candidate for nerve-sparing surgery—which means probability of impotence. Therefore, the presence of ECE bears heavily on treatment approach and strategy. As the UCLA team writes, “accurate ECE identification is crucial to assure oncological efficacy [cancer control] and functional outcomes [continence and potency] for RP patients.”^[i]

The problem is, how can a surgeon know in advance which patients have ECE? This is where Unfold AI comes in.

A test to compare best ECE predictors

The study team began their test with a pool of 241 PCa patients who were treated by RP. Before surgery, all had multiparametric MRI (mpMRI) prior to treatment; their images were interpreted by a radiologist who identified suspicious lesions and evaluated the risk of ECE by a 1-5 Likert scale (some patients also had PSMA PET scans done). All patients then had targeted-plus-systematic fusion-guided biopsies. When they then had RP, a pathologist examined the specimens and identified actual ECE status and locations. The pathology results were then the true basis for comparison with pre-surgery ECE prediction methods.

From this pool, the study team selected 147 cases based on criteria designed to “ensure data quality, a clinically relevant patient population, and compatibility with Unfold AI.”^[ii] The goal was to compare ECE assessment among 5 conventional methods (MRI Likert scores, capsular contact length of MRI-visible lesions, PSMA T stage, Partin tables, and the “PRedicting ExtraCapsular Extension” (PRECE) nomogram) and Unfold AI. Can you guess which method proved most accurate?

In the words of the study team, “Unfold AI accurately predicted ECE risk, outperforming conventional methodologies.” In fact, it “notably improved” ECE prediction in the specific prostate regions that would most influence a surgeon’s decision regarding nerve-sparing approaches. Thus, the team concluded, “By enhancing PCa staging and risk stratification, AI-based cancer mapping may lead to better oncological and functional outcomes for patients.

Thanks to the UCLA study, another AI resource can be added to the ever-advancing toolkit that promises to improve PCa clinical results and patient quality of life. Stay tuned for future AI blogs.

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] Priester A, Mota SM, Grunden KP, Shubert J, Richardson S, Sisk A, et al. Extracapsular extension risk assessment using an artificial intelligence prostate cancer mapping algorithm. BJUI Compass. 2024. https://doi.org/10.1002/bco2.421
[ii] Ibid.