AI model validation on LCS data

Work package three

The Optellum Virtual Nodule Clinic AI model is being trialled on Lung Cancer Screening (LCS) data, and an automated workflow is being developed to support implementation into clinical practice.

LCS project objectives

  • Model development and integration: where the existing algorithm is integrated into the LCS workflow to include automated detection and quantification.
  • Integration of clinical risk parameters and model calibration for UK lung screening population. The model will be extended by including the clinical risk parameters and Optellum will calibrate the resulting model for the UK LCS population to identify the threshold values entry into the various pathways.
  • Clinical Data Collection and Prospective Validation: the system developed earlier in the work package will be rolled out to the participating trusts and will undergo prospective clinical and health economic validation for every subsequent screening patient. Roche Diagnostics will work to ensure the enablement of the data models for point-of-care software deployment and utilization.

AI algorithm outputs are being validated using the same trial protocol currently in use for testing Optellum’s algorithm in an incidental pulmonary nodule NHS setting (IDEAL). This will lead to easy adoption into the NHS if the trials are successful.

Work package 3 is led by Prof Anand Devaraj and Dr Arjun Nair. Optellum is the industry partner.