This post refers to a newly published paper involving members of the van der Schaar Lab inside CCAIM. Find out more here.

In their paper, Changhee Lee, Alexander Light, Evgeny Saveliev, Mihaela van der Schaar, and Vincent Gnanapragasam present the first machine learning application for a dynamic risk prediction and temporal clustering of continuous data. This offers the opportunity to inform personalised follow-ups in prostate cancer active surveillance patients. Their algorithm, surpassing the current standards significantly, has the potential to be developed into a useable clinical tool.
Evgeny Saveliev, PhD Student at the van der Schaar Lab has created a short comprehensive video to explain the approach in a nutshell:
To learn more about the model, using cutting-edge machine learning (deep learning) to cut out assumptions and constraints of traditional statistical methods, you can discover an extensive explanative post on the van der Schaar Lab website here:
To illustrate the potential utility of their work, the authors provide an interactive web app, demonstrating a possible future clinical platform, which we encourage the reader to try out. More information about this can be found following the link above.
The paper is accessible here:
Changhee Lee, Alexander Light, Evgeny Saveliev, Mihaela van der Schaar, Vincent Gnanapragasam
Abstract
Active Surveillance (AS) for prostate cancer is a management option that continually monitors early disease and considers intervention if progression occurs. A robust method to incorporate “live” updates of progression risk during follow-up has hitherto been lacking.
To address this, we developed a deep learning-based individualized longitudinal survival model using Dynamic-DeepHit-Lite (DDHL) that learns data-driven distribution of time-to-event outcomes. Further refining outputs, we used a reinforcement learning approach (Actor-Critic) for temporal predictive clustering (AC-TPC) to discover groups with similar time-to-event outcomes to support clinical utility.
We applied these methods to data from 585 men on AS with longitudinal and comprehensive follow-up (median 4.4 years). Time-dependent C-indices and Brier scores were calculated and compared to Cox regression and landmarking methods. Both Cox and DDHL models including only baseline variables showed comparable C-indices but the DDHL model performance improved with additional follow-up data. With 3 years of data collection and 3 years follow-up the DDHL model had a C-index of 0.79 (± 0.11) compared to 0.70 (± 0.15) for landmarking Cox and 0.67 (± 0.09) for baseline Cox only. Model calibration was good across all models tested. The AC-TPC method further discovered 4 distinct outcome-related temporal clusters with distinct progression trajectories. Those in the lowest risk cluster had negligible progression risk while those in the highest cluster had a 50% risk of progression by 5 years.
In summary we report a novel machine learning approach to inform personalised follow-up during active surveillance which improves predictive power with increasing data input over time.