AI has the potential to bring standardized quality to global health care, but it will only gain widespread acceptance when it can show that it can also increase the specificity of diagnostics. Increasing specificity justifies artificial intelligence, reduces false positive diagnostics, leads to less burden for the patient, avoids unnecssary therapies and at the same time increases diagnostic efficacy for the lab. In order to demonstrate those effects, large-scale epidemiological studies with large patient numbers are necessary. We have built a globally accessable, cloud-based software platform supporting the easy building and management of large quanities of pathological image data as well as the application and validation of artificial intelligence algorithms. The first application arising from this platform is CYTOREADER.
CYTOREADER is our first cloud-based AI based system. It has, at present, been evaluated in epidemiological studies with over 50.000 cervical cancer patients with leading US institutions. CYTOREADER allows the automatic or assisted reading of dual-stain (p16/ Ki67) based cytology screening for cervical cancer. It is currently for research use only. In independent research studies of the US-NCI it has shown in full automatic mode a 10% increased specificity compared to human reading. It has the potential to broadly improve the women's health care situation in relation to cervical cancer.
Cervical cancer screening is currently in transition from conventional PAP screening to a combination of genetic HPV testing and biomarker-based (Ki67 / p16), dual stain (DS) cytology. PAP-testing has a well-documented subjectivity and a relatively low sensitivity and reproducibility. This results in a high workload for performing cytology involving rescreening of negative cytology results and other laborious quality controal procedures. It also requires retesting women with negative cytology results at short intervals. This can be overcome by manual DS testing, however, the manual evaluation of DS cytology is still subjective and time-consuming and it requires training and continuous quality control. Automated DS evaluation directly addresses these limitations by providing a completely objective cervical cancer screening approach, improving efficiency and reducing harms and costs related to false-positive screening results.
STCMED is the Steinbeis Transfer Center for Medical Systems Biology, which is part of the Steinbeis GmbH & Co KG for technology transfer in the large network of the Steinbeis Foundation. The Steinbeis Foundation is a non-profit institution founded by the German state of Baden-Württemberg promoting the transfer of technology and knowledge between universities and industry. We are an international team of experts creating and managing large studies based on digital pathology images for the creation of artificial intelligence based assisted diagnostics. To this purpose we are closely collaborating with Heidelberg University and the US National Cancer Institute (NCI), Division of Cancer Epidemiology and Genetics (DCEG).