Deep learning colorectal H&E tumour detection algorithms seamlessly integrated in the clinician’s workflow though our LIMS (Laboratory Information Management System)
The core focus of our innovation is a fully integrated pathology workflow. From LIMS to digital pathology to AI, we aim to achieve our company goal of Improving Wellbeing Through Innovation and thus improving the efficiency of clinicians with less time spent on the mundane, ultimately improving patient outcomes through faster and more accurate diagnosis.
To this end, we have developed a world class product TissueMark to offer deep learning tissue detection algorithms to clinicians working on colorectal cancer. This algorithm automatically identifies and detects tumour tissue in colorectal H&E slides. In addition, we are expanding automated detection capabilities to distinguish benign and malignant colon polyps to drive efficiencies in Bowel Cancer Screening in the UK. A large part of this project is the explainability of what the algorithm is detecting to augment the robustness in a real-world environment. Our approach is based on fuzzing, whereby random inputs are applied to the images to try to confuse the algorithm. This will enable us to identify the salient parts of the input data and the different feature layers, which will help to explain the behaviour of the networks as well as reveal the algorithm’s less robust input features which will then help us identify solutions to improve the training data.
One of the unique propositions Cirdan can offer is integration of new AI tools with our ULTRA LIMS (Laboratory Information Management System) that is currently deployed in hospitals and trusts across the UK, Australia, Canada, and others. This puts novel AI algorithms for tissue detection and biomarkers scoring directly into the hands of pathologists quickly, giving them seamless access to AI directly accessible through the diagnostic reporting in the LIMS.
NDA is required before full disclosure of innovations.
The innovation will drive integrated AI reporting in colorectal pathology, through the identification of cancer in colon biopsies and resections specimens and the automated discrimination of polyps as part of the colorectal cancer screening programme. The ability to run AI within the Ultra LIMS is essential for getting this powerful AI into the hands of pathologists, driving efficiencies in diagnostic colon pathology and overcoming the adoption issues that exist in pathology AI today.
Though a partnership we could offer not only the deep learning algorithms already developed but also the training, test, and validation data (20,000 Whole Slide Images from different body sites), we also have multiple data partnerships for access to images beyond that. Additional to the data opportunities we have many long-standing connections with leading pathologists and researchers across the UK that could be used to help any joint venture.
We are also working with researchers from Queen's University Belfast who have been involved in a previous project on gross tumour definition in non-small lung cancers for radio therapy treatment planning. They also have been involved in the segmentation, detection and tracking of centrosomes, telomeres, cells, and focal adhesions in confocal microscopy imagery with extensive experience applying deep learning to video surveillance, financial data and cyber security and have developed machine learning technology up to TRL5/6.
The following applications apply:
- Automated detection of invasive colorectal cancer in H&E stained tissue sections
- The annotation of tissue samples for molecular macrodissection
- The assessment of tissue cellularity and % tumour cells
- The automated assessment of tissue sample quality (tumour cell content) for molecular testing in colorectal cancer
- The automated annotation of tumour for biomarker evaluation
- Deep learning models to detect molecular anomalies in colorectal tissues from H&E
- The training of pathologists in colorectal tissue interpretation
- The automated annotation of colorectal tissues in biobanking
- The use of AI algorithms as foundation for other tissue biomarkers
- The integration of AI within a powerful LIMS to create seamless workflow and accelerate adoption of AI