anonmaths / tumour_sde_model

Variational inference in an SDE tumor growth model using a recurrent neural network

---

The main resource is: https://arxiv.org/abs/1802.03335/

---

The code is based on: https://github.com/Tom-Ryder/VIforSDEs/

---

The source of the dataset: http://clincancerres.aacrjournals.org/content/18/16/4385

---

A similar model specification: https://arxiv.org/abs/1607.02633

---

The model is given by the system of SDEs:

where

is the number of observations for subject i,

maps the ith patient to their treatment group index.

The noise is assumed to be distributed according to

• 

which may equivalently be written as

• 

The assumed prior distributions:

• 

where and are treatment group dependent hyperparameters, also

• 
• 
• 
•