Novel Automated Three-Dimensional Surgical Planning Tool and Magnetic Resonance Imaging/Ultrasound Fusion Technology to Perform Nanoparticle Ablation and Cryoablation of the Prostate for Focal Therapy.

Purpose: Although MRI/ultrasound fusion has been primarily used to assist in the diagnosis of prostate cancer, this technology can also be used to focally treat localized prostate cancer. We present one case of nanoparticle-directed ablation and two cases of cryoablation to focally treat prostate tumors. Patients and Methods: Three patients underwent MRI/ultrasound fusion transperineal prostate biopsies to confirm low- to intermediate-risk prostate cancer. The MRI lesions correlated with the biopsy-proven disease. Pelvic MRI segmentation was performed with DynaCAD 5.0 workstation. The MRI lesion including a 6 to 10 mm margin, prostate, bladder, urethra, urethral sphincter, rectum, and pubic bone were segmented. MRI/ultrasound fusion was performed with the novel Philips UroNav 4.0 system. Lesions were treated with focal nanoparticle ablation or focal cryoablation. Results: A 69-year-old man with a right posterior medial peripheral zone lesion positive for Gleason grade group (GG)3 cancer was treated with focal nanoparticle ablation. The UroNav 4.0 system reported 100% ablation of the segmented tumor and 94% of the 6 to 10 mm margin at the end of the case. A 68-year-old man with a left anterior fibromuscular stroma lesion positive for Gleason GG2 cancer and a 71-year-old man with a right peripheral zone posterior lateral lesion positive for Gleason GG1 cancer were treated with focal cryoablation. The UroNav 4.0 system reported 100% ablation of the segmented tumor and 82% of the 6 to 10 mm margin at the end of the case. Conclusion: Observation of the prostate tumor(s), surrounding critical structures, and pelvis in three dimensions (3D), along with the anticipated ablation zone, is one of the challenges of pelvic surgery and percutaneous ablation. The DynaCAD 5.0 Urology system can create an auto-segmented 3D rendering of critical structures and the tumor(s), as well as observation and quantification of the anticipated ablation coverage, to facilitate preoperative planning of needle placement. ClinicalTrials.gov nos.: NCT02680535 and NCT04656678.

Exploiting Rules to Enhance Machine Learning in Extracting Information From Multi-Institutional Prostate Pathology Reports.

PURPOSE: Literature on clinical note mining has highlighted the superiority of machine learning (ML) over hand-crafted rules. Nevertheless, most studies assume the availability of large training sets, which is rarely the case. For this reason, in the clinical setting, rules are still common. We suggest 2 methods to leverage the knowledge encoded in pre-existing rules to inform ML decisions and obtain high performance, even with scarce annotations. METHODS: We collected 501 prostate pathology reports from 6 American hospitals. Reports were split into 2,711 core segments, annotated with 20 attributes describing the histology, grade, extension, and location of tumors. The data set was split by institutions to generate a cross-institutional evaluation setting. We assessed 4 systems, namely a rule-based approach, an ML model, and 2 hybrid systems integrating the previous methods: a Rule as Feature model and a Classifier Confidence model. Several ML algorithms were tested, including logistic regression (LR), support vector machine (SVM), and eXtreme gradient boosting (XGB). RESULTS: When training on data from a single institution, LR lags behind the rules by 3.5% (F1 score: 92.2% v 95.7%). Hybrid models, instead, obtain competitive results, with Classifier Confidence outperforming the rules by +0.5% (96.2%). When a larger amount of data from multiple institutions is used, LR improves by +1.5% over the rules (97.2%), whereas hybrid systems obtain +2.2% for Rule as Feature (97.7%) and +2.6% for Classifier Confidence (98.3%). Replacing LR with SVM or XGB yielded similar performance gains. CONCLUSION: We developed methods to use pre-existing handcrafted rules to inform ML algorithms. These hybrid systems obtain better performance than either rules or ML models alone, even when training data are limited.