CLinical Evaluation Of a comPuter Algorithm To Report BreAst cAncers

This study follows on from work conducted over the last 4 years focusing on developing computer-aided diagnostic algorithms in histopathology, with a focus on the detection and typing of breast cancer. The algorithm extracts information from the same digital image of a breast biopsy glass slide as used by pathologists to make a diagnosis. The algorithm can identify the presence of invasive cancer, grade it (i.e. how aggressive it looks) and determine its molecular subtype, all from the same image and in a single step. Whilst a pathologist can identify invasive cancer and grade it on a biopsy specimen, grading is inconsistent across reporting clinicians (i.e. prone to inter-observer variability) and only about 80% aligned with the 'true' grade of the tumour obtained at surgery. Moreover, additional time-consuming and expensive laboratory tests are required for molecular subtyping. They are however essential for informing oncologists of the most effective/targeted therapy for the patient in each case. The aim of this study is to perform a final real-life acid test of the technology in the clinical environment separately but in parallel with routine diagnostic services. If successful, the technology will have acquired a sufficient body of data to convince end-users and regulatory bodies of its merit and will effectively make it ready for deployment in the clinical environment. In that domain, this diagnostic solution health economics assessment will run alongside the clinical evaluation of the algorithm to analyse and compare cost and time-related factors between the routine clinical service and the algorithm workflow. This will provide further information required by regulatory bodies to aid the consideration of utilising the technology within current health systems. Furthermore, follow-up data will be collected at 5 and 10 years to investigate patient prognosis and treatment response as well as future potential applications of the technology.