Medial EarlySign’s research ranges from early prediction of CRC and epilepsy seizures to data-driven diagnostic tools for ICU, personalized and outcome based interpretation of lab results, stability predictors for PCI units, and more. Our model development adheres to rigorous clinical methodologies through use of millions of data points from around the world.
For example, ColonFlag’s CRC model for identification of individuals at high risk of having colorectal cancer (CRC) was developed using data of over 605,000 Maccabi Healthcare Services (MHS) patients aged 40+. MHS is the second largest HMO in Israel, with approximately 2 million patients. Data extraction, model development and analysis were all performed according to an Ethics Committee approved protocol. All patients’ records were anonymized and de-identified prior to analysis. This cohort, which included 3,135 CRC cases, was randomly divided into a derivation (training) dataset (80% of the data) and a separate validation dataset (20% of the data). The model was further validated using de-identified data of 25,613 individuals above age 40 (5,061 CRC cases), from The Health Improvement Network (THIN), a UK primary care database. This database represents an external independent validation set.
ColonFlag is further investigated and validated in relation to CRC, Advanced Adenoma and precancerous polyps in various clinical studies in leading medical and health system institutions and CRC screening centers around the world, involving key opinion leaders in gastroenterology and public health. The studies sites are located in Israel, North America (United States and Canada), Europe and Asia, thus representing diverse populations.
Study designs, both retrospective and prospective, range from case-control designs to a cohort design of more than 1.5 million individuals.
Additional models for cancers (upper GI and Lung), epilepsy seizure prediction, and data diagnostic tools for ICU predication are being designed and clinically validated in Israel, United State and Asia.
Download ● JAMIA Publication