298

---

Click the image below to view the full size version of this cover.

Issue 298: Adaptive Therapy has been developed as an alternative treatment scheduling paradigm, aiming to control rather than try to cure late-stage cancers by including breaks in treatment to resensitise the tumour to the applied therapeutic. Previous adaptive approaches have employed a 'one size fits all' approach to scheduling these breaks, applying the same algorithm to all patients despite their widely different tumour dynamics. By integrating mathematical modelling and machine learning into the decision-making process, we can tailor these schedules to individual patients, as they follow a unique trajectory under treatment. Shown are three such response trajectories, spiralling out from a common centre as the patients' response to treatment diverges over time, and the scheduling adapts to this. Each trajectory alternates between crosses that represent the patient's clinical data for a single treatment cycle, from which we fit a mathematical model (given by the solid line), and use this model to drive a machine learning framework (1s and 0s) which arrives at treatment recommendation for the next cycle, before we repeat this process. Based on the paper: Mathematical Model-Driven Deep Learning Enables Personalized Adaptive Therapy published in Cancer Research.