Title: Integrating biomechanistic models and machine learning classifiers for personalized prediction of prostate cancer progression and radiotherapy response

Abstract: The current clinical protocols to manage prostate cancer (PCa) enable the detection and successful treatment of these tumors at an early stage. However, recent studies suggest that many PCa patients are being overtreated, and hence prone to potential treatment side-effects (e.g., incontinence, impotence) that can adversely impact their quality of life without improving longevity. Furthermore, undertreatment of PCa is another important clinical challenge, as it may lead to rapid growth of aggressive tumors, treatment failure, and reduced survival. The overtreatment and undertreatment of PCa have the same origin: the limited individualization and observational nature of the clinical management of these tumors. In this talk, I propose to address these critical, unresolved issues by using patient-specific forecasts of PCa growth and treatment response, along with hybrid classifiers that take biomechanistic inputs to predict the occurrence of clinical events of interest. I will present the application of this predictive framework in two scenarios where longitudinal data are collected as part of the standard-of-care management of PCa: active surveillance of lower-risk tumors before primary treatment, and the post-treatment monitoring of patients after radiotherapy. For each application, I will show how a biomechanistic model can be built, calibrated, and validated to obtain personalized predictions of tumor growth and therapeutic response. Then, logistic classifiers will be trained with biomechanistic model outputs to identify tumors progressing towards higher-risk disease during active surveillance or developing a recurrence after radiotherapy. Finally, although further development and validation over larger cohorts are needed, I will posit that the technologies presented in this talk can contribute to advance the observational, population-based standards in clinical oncology towards a predictive, personalized paradigm.

About the speaker: https://www.glorenzophd.com