Mp30-04 Development of a Computer AIded Diagnosis System Using Multiparametric Transrectal Ultrasound for the Localization of Clinically Significant Prostate Cancer

You have accessJournal of UrologyImaging/Uroradiology II (MP30)1 May 2024MP30-04 DEVELOPMENT OF A COMPUTER AIDED DIAGNOSIS SYSTEM USING MULTIPARAMETRIC TRANSRECTAL ULTRASOUND FOR THE LOCALIZATION OF CLINICALLY SIGNIFICANT PROSTATE CANCER Daniel L. van den Kroonenberg, Auke Jager, Arnoud W. Postema, Katelijne de Bie, Marinus J. Hagens, Hessel Wijkstra, Peet T. G. A. Nooijen, Hans van der Linden, Joost de Baaij, Jean-Peal A. van Basten, Pim J. van Leeuwen, Henk G. van der Poel, Harrie P. Beerlage, Massimo Mischi, and Jorg R. Oddens Daniel L. van den KroonenbergDaniel L. van den Kroonenberg , Auke JagerAuke Jager , Arnoud W. PostemaArnoud W. Postema , Katelijne de BieKatelijne de Bie , Marinus J. HagensMarinus J. Hagens , Hessel WijkstraHessel Wijkstra , Peet T. G. A. NooijenPeet T. G. A. Nooijen , Hans van der LindenHans van der Linden , Joost de BaaijJoost de Baaij , Jean-Peal A. van BastenJean-Peal A. van Basten , Pim J. van LeeuwenPim J. van Leeuwen , Henk G. van der PoelHenk G. van der Poel , Harrie P. BeerlageHarrie P. Beerlage , Massimo MischiMassimo Mischi , and Jorg R. OddensJorg R. Oddens View All Author Informationhttps://doi.org/10.1097/01.JU.0001009416.90901.7b.04AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Computer-aided diagnostic (CAD) systems based on artificial intelligence (AI) have great potential to meet the growing demand for image processing. Our aim is to develop a CAD system using three-dimensional multiparametric transrectal prostate ultrasound (3D mpUS) to localize clinically significant prostate cancer (csPCa). METHODS: We included 252 "positive" patients with csPCa (Gleason Group≥2) scheduled for radical prostatectomy (RP) and 83 "negative" patients with suspicion for PCa but with negative MRI (PIRADS≤2) or negative systematic biopsies, for training the CAD system. Patients underwent 3D mpUS, consisting of 3D grayscale, 4D contrast-enhanced ultrasound (CEUS), and 3D shear wave elastography (SWE) using automated acquisition sequence. The histopathology of whole-mount RP specimen of the positive patients provided the ground truth for CAD training and was registered to the mpUS images using custom software. Quantitative features were extracted from SWE, B-Mode and CEUS establishing relevant imaging biomarkers for PCa. These were optimally combined by a dedicated AI model that was trained for voxel-based detection of csPCa based on the ground truth. Training and validation of the AI model was performed using 5-fold cross validation repeated over 3 different random splits of the dataset. We assessed diagnostic accuracy of the developed CAD for csPCa by comparing ground truth voxels with CAD predicted voxels. RESULTS: Based on the input of 264 patients, a sensitivity of the CAD of 0.82 and specificity of 0.82 is reached on a voxel level for the detection of csPCa, with a AUROC of 0.90. The negative predictive value and positive predictive value were 0.65 and 0.92. The dataset is not a true representation of the general population, as it holds a prevalence of 70% of csPCa. By developing the CAD using negative patients, they could have been false negative on MRI or false negative on systematic biopsies. CONCLUSIONS: The developed CAD system based on mpUS shows accurate voxel-level detection of csPCa. We now explore a lesion level validation and are setting up a prospective clinical validation trial to further investigate the diagnostic performance of this CAD system and its clinical applicability. Download PPT Source of Funding: European Innovation Council (EIC) transition Angiogenesis Analytics © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e491 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Daniel L. van den Kroonenberg More articles by this author Auke Jager More articles by this author Arnoud W. Postema More articles by this author Katelijne de Bie More articles by this author Marinus J. Hagens More articles by this author Hessel Wijkstra More articles by this author Peet T. G. A. Nooijen More articles by this author Hans van der Linden More articles by this author Joost de Baaij More articles by this author Jean-Peal A. van Basten More articles by this author Pim J. van Leeuwen More articles by this author Henk G. van der Poel More articles by this author Harrie P. Beerlage More articles by this author Massimo Mischi More articles by this author Jorg R. Oddens More articles by this author Expand All Advertisement PDF downloadLoading ...