Menu
Menu
Cardiovascular diseases (CVDs) are the leading cause of death globally, with 17.9 million deaths annually. Syndesis has partnered with Insybio, an AI pioneer in biomarker discovery and medical predictive analytics, to address this enormous global health problem.
We focused on two specific areas, congestive heart failure (CHF) and hypertrophic cardiomyopathy (HCM). Congestive heart failure (CHF), or heart failure, is a long-term condition in which the heart cannot pump blood well enough to meet the body’s needs. Hypertrophic cardiomyopathy (HCM) is considered a particularly unfavourable disease complication associated with high mortality and frequently requiring heart transplant. Doctors are challenged when faced with patients suffering from CHF or HCM, in ensuring the optimal therapeutic protocol is applied to reduce the risk of dangerous complications or death.
Existing clinical scores (Qrisk2, SCORE, Framingham) perform poorly in predicting 5-year mortality and short-term outcome (<60% Area Under the Curve, AUC) in both disease areas. Artificial Intelligence can provide significant support in such clinical decisions by predicting outcomes for patients given a certain protocol.
Insybio used a highly curated cardiovascular data set on Syntium, Syndesis Health’s data platform of anonymized clinical data, to develop two Precision Medicine Tools based on Artificial Intelligence that predict risk and recommend changes in the therapeutic protocol of CHF and HCM patients that reduce predicted risk.
The AI tools – “PrecisionCHF” and “PrecisionHCM” – were trained on 1044 patients hospitalized with Congestive Heart Failure (CHF) and 255 Hypertrophic cardiomyopathy (HCM) patients with full clinical annotation and outcomes mined from Syntium.
Following training on the data the AI tools performed impressively well: PrecisionCHF was able to predict 5-year mortality for CHF with 75% AUC. PrecisionHCM was able to predict short term outcome (i.e. mortality until discharge) with 97% AUC. Using the AI tools 37% of CHF and 28% of HCM patients at risk can substantially reduce their risk (>5% reduction) with the optimization of their medication plan. Moreover, 12% of the rest of CHF and 11% of the rest of HCM who are not at risk can further reduce their risk by at least 5%.
The collaboration between Syndesis and Insybio has demonstrated the enormous value of training AI pipelines for medical applications using relatively “small” but highly curated and validated clinical data on Syntium.