Attention: Restrictions on use of AUA, AUAER, and UCF content in third party applications, including artificial intelligence technologies, such as large language models and generative AI.
You are prohibited from using or uploading content you accessed through this website into external applications, bots, software, or websites, including those using artificial intelligence technologies and infrastructure, including deep learning, machine learning and large language models and generative AI.

FOCAL THERAPY Artificial Intelligence for Prostate Cancer Localization and Focal Therapy Patient Selection

By: Wayne G. Brisbane, MD, UCLA Urology, Los Angeles, California | Posted on: 09 Jun 2023

Prostate cancer is the most common solid malignancy diagnosed among men in the United States.1 Most men are diagnosed with localized disease. Focal therapy for prostate cancer is an emerging treatment option for men that treats the prostate tumor while sparing benign tissue. As prostate cancer is often multifocal,2 focal therapy assumes that a single index lesion is responsible for prostate cancer metastasis. Index lesion ablation would thus prevent future metastasis without collateral damage to adjacent structures critical for sexual and urinary function.3

Successful focal therapy depends on 3 features: (1) appropriate patient selection, (2) precise localization of the index lesion, and (3) effective delivery of an ablation energy.

Most recent advancements in focal therapy have increased the armamentarium of ablative energy options. Although for a decade, patient selection and index lesion localization have relied on tumor grade and MRI, respectively. However, a recent Food and Drug Administration (FDA)–approved technology is promising to improve patient selection and index lesion localization for patients interested in focal therapy.

Avenda Health, a technology startup with roots in the Departments of Urology and Engineering at University of California, Los Angeles (UCLA), received FDA clearance for its “Unfold AI” software. Unfold AI is a machine learning algorithm that receives inputs of patient clinical features, MRI images, and tracked prostate biopsy cores to deliver a 3-dimensional cancer probability map (see Figure).

Figure. Example case of Unfold AI. The algorithm accepts patient clinical features, MRI images with radiologist contours (A) and pathology with tracked biopsy core locations (B). C, Cancer probability map is displayed as a heat map delineating the most likely position of cancer. The recommended tumor margin is outlined in pink (C). The margin can be adjusted by the provider, and the software will subsequently display the adjusted confidence of encapsulating the entire tumor (example not shown). D, Final pathology for the Unfold AI example. Note that the MRI return on investment and positive biopsy cores (B, yellow cores) do not well delineate the actual tumor volume; however, the cancer is well delineated by the heat map (C) and would be completely treated with the suggested margin.

Further, a patient-specific treatment margin is generated, which balances treating all clinically significant cancers against minimizing side effects. For clinicians who provide education to a range of focal therapy candidates, the software’s graphical interface intuitively demonstrates where focal therapy is reasonable vs overreaching.

In data-pending publication, the use of Unfold AI seems to identify tumor margins more accurately than the current standard of care. Similar to MRI, the 3D cancer map can be fused to focal therapy modalities to direct intraprostatic treatment margins. It is hopeful that Unfold AI will improve consistency in lesion identification and ablation.

To be sure, improvements in focal therapy are needed. Taking the simple definition of focal therapy failures—residual clinically significant prostate cancer following focal therapy—failure rates range from 24% to 41% at 6 months.4,5 Until now, focal therapy failure has been multivariable, with incremental deviations from all 3 elements of success. Unfold AI may offer improved patient selection and index lesion localization, thus allowing future research to focus on improving energy delivery to the index lesion.

However, there is moderate cause for concern. Unfold AI was developed at UCLA and then validated at Stanford. Both institutions are high-volume centers of excellence utilizing a single transrectal fusion platform (Artemis); it is unclear whether their results are duplicated easily. Additionally, the technology pivots on the assumption that a radiologist correctly interprets the MRI, and then the ultrasound and MRI are fused accurately—both are variable.6 Finally, it is unclear whether the ability to precisely predict the index tumor+margin will overcome failures related to multifocal disease.

Yet, there is also cause for optimism. With a wide breadth of ablative options available, Unfold AI represents the first technology to improve tumor localization and patient selection. After using the technology to create the Figure, I am hopeful that Unfold AI will be to intraprostatic staging what prostate-specific membrane antigen positron emission tomography/CT has been for extraprostatic staging. Additionally, if the accuracy proves founded, there are likely potential applications for surgery, radiation therapy, and patient decision-making.

  1. Siegel RL, Miller KD, Wagle, NS, Ahmedin J, Jemal A. Cancer statistics, 2023. CA Cancer J Clin. 2023;73(1):17-48.
  2. Johnson DC, Raman SS, Mirak SA, et al. Detection of individual prostate cancer foci via multiparametric magnetic resonance imaging. Eur Urol. 2019;75(5):712-720.
  3. Ahmed HU. The index lesion and the origin of prostate cancer. N Engl J Med. 2009;361(17):1704-1706.
  4. Aker MN, Brisbane WG, Kwan L, et al. Cryotherapy for partial gland ablation of prostate cancer: oncologic and safety outcomes. Cancer Med. 2023;10.1002/CAM4.5692.
  5. Mortezavi A, Krauter J, Gu A, et al. Extensive histological sampling following focal therapy of clinically significant prostate cancer with high intensity focused ultrasound. J Urol. 2019;202(4):717-724.
  6. Stabile A, Giganti F, Kasivisvanathan V, et al. Factors influencing variability in the performance of multiparametric magnetic resonance imaging in detecting clinically significant prostate cancer: a systematic literature review. Eur Urol Oncol. 2020;3(2):145-167.