Mp25-08 Software to Determine Extent of Tumor Margins in Focal Therapy of Prostate Cancer (Unfold-AI®)

You have accessJournal of UrologyProstate Cancer: Localized: Ablative Therapy I (MP25)1 May 2024MP25-08 SOFTWARE TO DETERMINE EXTENT OF TUMOR MARGINS IN FOCAL THERAPY OF PROSTATE CANCER (UNFOLD-AI®) Wayne G. Brisbane, Alan Priester, Mark T. Topoozian, Anissa V. Nguyen, Merdie K. Delfin, Samantha R. Gonzalez, and Leonard S. Marks Wayne G. BrisbaneWayne G. Brisbane , Alan PriesterAlan Priester , Mark T. TopoozianMark T. Topoozian , Anissa V. NguyenAnissa V. Nguyen , Merdie K. DelfinMerdie K. Delfin , Samantha R. GonzalezSamantha R. Gonzalez , and Leonard S. MarksLeonard S. Marks View All Author Informationhttps://doi.org/10.1097/01.JU.0001008692.26556.39.08AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Margins of prostate cancer (PCa) are poorly defined by MRI, but margins must be treated to ensure a complete ablation. Herein we test a recently FDA-cleared software, Unfold-AI®, to improve the definition of tumor margins and guide focal therapy. METHODS: Unfold-AI®, a PCa margin-prediction software based on artificial intelligence (AI), was described by Priester (Eur.Urol. O.S. vol. 54, page 20-27, 2023). It is derived from biopsy and clinical information of 875 men undergoing 1145 MRI-guided biopsies (13k cores, 128k MR images). The Unfold-AI® output is a patient-specific encapsulation confidence score (ECS) estimating the probability of complete tumor ablation when recommended margins are followed. ECS values from 0 to 1 (Figure 1). are directly related to the probability of focal therapy success.We retrospectively applied Unfold-AI® in 118 treatment-naïve men in a trial of hemi-gland cryotherapy (CRYO) for PCa >GG2 (NCT03503643). Median age was 68 yr (IQR 63-72); 67% were Caucasian, 5% Black, 28% other; PSA 6.8 ng/ml (IQR 4.7-10.3); prostate volume 44cc (IQR 32-56). MRI-guided biopsy (MRGB), targeted + systematic, was performed before and 6 months after CRYO. To determine predictive value, the ECS score and standard PCa variables were compared against the presence or absence of >GG2 in post-CRYO MRGB. ROC analysis was used to assess the significance of the metrics in predicting CRYO outcomes (Table 1). RESULTS: Baseline GG was 66.8% GG2, 28.0% GG3, and 5.2% GG4. Treatment success (no >GG2 on MRGB) was 77.7% and was not related to baseline GG. Univariate ROC analysis (Table 1) showed that max cancer core length, % Gleason pattern 4, and ECS were predictive of success. However, in a multivariate logistic regression, only ECS≥0.7 remained as a predictor of success (aOR 4.7, 1.7-12.9). Sensitivity and specificity for ECS≥0.7 was 70% (61-79) and 68% (51-85), respectively. CONCLUSIONS: In predicting focal therapy success, tumor margins -- as determined here by the ECS of Unfold-AI® --may be more important than other baseline parameters (e.g., GG). Download PPT Source of Funding: NCI-RO1CA218547 © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e407 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Wayne G. Brisbane More articles by this author Alan Priester More articles by this author Mark T. Topoozian More articles by this author Anissa V. Nguyen More articles by this author Merdie K. Delfin More articles by this author Samantha R. Gonzalez More articles by this author Leonard S. Marks More articles by this author Expand All Advertisement PDF downloadLoading ...