This project will result in an AI algorithm that is clinically useable and ready for FDA submission
Area of work
Malignant Pleural Mesothelioma
Size of data set
Professor Kevin Blyth
Number of NHS partners1
Our Predict Meso project is focused on Malignant Pleural Mesothelioma (MPM) which is currently an incurable cancer that develops many decades after inhalation of asbestos dust. Few treatment options exist, and most patients die within a year. Pleural disease is difficult to accurately measure, and it is known that reporters are inconsistent in their assessment of disease volume. The standard form of disease measurement in oncology, RECIST – Response Evaluation Criteria In Solid Tumours – has been modified for MPM, RECIST 1.1, but is still regarded as unsatisfactory.
Blyth et al have shown that computer-based algorithms that semi-automate the critical step of volumetric tumour segmentation out-performs traditional staging when based on MRI (Tsim et, Lung Cancer 2020). It is now necessary to improve upon these algorithms, using large image data sets linked to histology, treatment and disease outcomes, making use of the developments available using Artificial Intelligence. This will then enable them to move from the research environment into clinical validation and then clinical practice.
The aim of this project is to further develop, and then validate an AI algorithm that may be used to measure the volume of pleural disease in patients with mesothelioma, and more accurately determine disease response. The study will collect data from UK centres looking after patients with malignant pleural mesothelioma, MPM. Using this data the artificial intelligence, AI, algorithms already developed by Professor Blyth and Canon Medical Ltd (Anderson et al, Proc 13th IJCBEEST: Bioimaging 2020) will be optimized using MPM CT data transferred from the NCIMI network.
The AI algorithm will automatically identify the tumour within the chest and outline it – known as segmenting the malignant pleural disease. The algorithm will automatically produce a volume of the total amount of tumour and will then be able to accurately determine the volume of tumour when the patient presents, and also accurately determine whether they are responding to their treatment.
No articles found