Periodic Reporting for period 2 - Contextflow (3D image-based search engine and triaging software for lung CTs to improve radiology workflows) Reporting period: 2021-09-01 to 2022-08-31 Summary of the context and overall objectives of the project Radiology is struggling. Exponentially increasing quantities of data make it difficult to find relevant information. Radiologists' workload is also rapidly increasing, exacerbated by the global radiologist shortage. What's more, new treatments require more complex diagnoses. When faced with a difficult case,radiologists must currently discuss with colleagues, consult reference books or guess search terms in text-based resources. This frustrating, time-consuming process leads to delays, missed findings and high overtime expense. contextflow develops deep learning-based tools to improve radiology workflows, saving time while increasing reporting quality. Lung diseases are particularly hard to diagnose; they are characterised by the combination and distribution of 40 anomalous patterns observed in lung CTs. contextflow software reduces the time radiologists spend searching for information, allowing for both faster and higher-quality diagnostics. Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far The project builds on existing contextflow technology, a 3D image-based search engine. This search engine can already detect 19 disease patterns, instantly linking a 3D image to reference cases with similar findings, case statistics and reference information necessary for differential diagnosis. During the project, we have developed methods to detect lung diseases, including COVID-19, based on the anomalous lung disease patterns; and identifying signatures indicative of diseases in medical data to support personalised treatment decisions. We have collected and manually annotated extensive lung CT datasets in order to carry out this development. We have also developed technology to follow changes in lung disease patterns of a single patient over time. A highlight of the dissemination activities is a series of videos explaining the contextflow technology. contextflow closed a €6.7 million Series A investment round in 2021. Results of the project are implemented in the contextflow ADVANCE Chest CT software. Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far) The contextflow system is currently able to detect 19 lung disease patterns (other systems can detect 1-3 patterns). Linking these disease pattern distributions to lung diseases results in a significant step forward in AI support for radiologists. A further breakthrough is the timeline analysis, which provides radiologists with tools to track lung disease progression in a patient over time. Through collaborations with PACS vendors (Picture Archiving and Communication Systems), contextflow stimulates the broad adoption of these techniques by radiology departments. Screenshot of the contextflow lung CT software
