
An online tool for breast cancer prognostication and decision support
Adjutorium is an online machine learning-based prognostication tool that enables patients and clinicians to predict survival rates and treatment benefits for early invasive breast cancer after breast-conserving surgery.
For women with early-stage breast cancer
Adjutorium predicts survival outcomes for patients with early breast cancer after breast conserving surgery. It can be used in consultation with specialists to decide whether adjuvant therapies are right for any given patient.


Designed to complement the expertise of clinicians
Adjutorium predicts the most likely outcomes for a patient based on current knowledge and data, but will never provide a 100% accurate prediction for any individual. Adjutorium’s predictions serve to complement discussions with specialists in a more personalised context.
State-of-the-art machine learning
Adjutorium makes predictions using a state-of-the-art automated machine learning model that was trained using historical data for patients.
The model requests some information about the patient, and uses a machine learning algorithm to predict the patient’s survival profile under different treatment options using data for similar women diagnosed with breast cancer in the past.


Validated using data for nearly 1 million women
Adjutorium was trained and externally validated on data for nearly 1 million women captured in the cancer registries of the UK and the US.
Developed by a multidisciplinary team of academics
Adjutorium was developed through by the van der Schaar Lab through a collaboration with Public Health England and a multidisciplinary team of academics from UCLA, the University of Cambridge, The University of Oxford, Queen Mary University, and The Alan Turing Institute.

Interested?
Read the paper about Adjutorium in Nature Machine Intelligence, try Adjutorium live via our web app, or explore the source code for AutoPrognosis, the AutoML model behind Adjutorium.
Note: the web app may take a few seconds to load, depending on the number of current users. Similarly, there may be a slight lag before the graph and predictions respond to changes in user inputs. Please give it a bit of time, and avoid refreshing the page.