Following our previous blog: ‘Unveiling the importance of Age in Oesophageal Cancer Treatment Decisions: An Interpretable Machine Learning Approach,’ our team have developed the model further to be able to predict treatment decisions and survival for oesophageal cancer patients. This happened at Southampton, based on data from (~1000 patients). Having collaborated with Oxford University Hospitals we also have an external dataset to be able to validate our model in a different environment.

This increases the generalisability of our findings. Preliminary results are encouraging, showing that our model performs consistently well across different institutions. Additionally, we have developed capabilities to predict survival probabilities for palliative care patients at various time points, which are seamlessly integrated into our user-friendly clinical interface.

Importantly, we have also been able to gauge UK wide interest in such a tool that might be implemented in the NHS. 75% of respondents to a national survey aimed at regular Oesophageal Cancer Multi Disciplinary Team members were supportive of such a Machine Learning (ML) based decision support tool. But this study also highlighted barriers to routine use of AI/ML in such a critical sector as health, with some experts concerned about transparency and over the model’s ability to adapt to changes in practice. These insights are crucial as we plan further enhancements to our tool, informed by a series of in-depth interviews with expert clinicians.

We are looking forward to conducting patient focus groups as this project concludes, ensuring that our tool is attuned to user preferences and needs.

Our work continues to receive accolades with further prizes such as the Williams Prize 2024 from the Surgical Research Society (Cambridge), as well as presenting our work at the latest RAI UK Health Panel meetings. We were thankful for being able to demonstrate our model at the TAS Showcase 2024 (pictured below), in sparking interest and discussions for collaboration with experts across multiple domains.

Multiple manuscripts are currently in preparation/under review and we are aiming for maximum impact with potential applicability to multiple types of cancer.

As this project continues, more avenues keep cropping up and we are excited to embark on the next steps of implementation of an AI/ML based tool to support decision making for our cancer patients across the UK.

Read more about their project: Reformist: Mirrored decision support fRamEwork FOR MultidISciplinary Teams in Oesophageal cancer here. 

Navamayooran (Nav) Thavanesan demonstrating the decision making software to Dr Tina Seabrooke
Navamayooran (Nav) Thavanesan demonstrating the oesophageal cancer decision making software to Dr Tina Seabrooke at the TAS Showcase 2024.