The deep learning algorithm which forms the heart of Deep Sepsis API has been developed for early detection of sepsis condition in critical care patients. Compared to traditional scoring mechanisms like SIRS, aSOFA, MEWS, NEWS, deepSepsis AI algorithm has much higher accuracy and specificity. The deep learning model was developed and validated with over 50,000 patient records and 32 million data points at Duke University, North Carolina.
Built using Multi-Gaussian Process Recurrent Neural Network (MGP-RNN) that takes input from each clinical value recorded from the time of admission to arrive at a risk score, every time.
Corroborates across millions of data points to predict the possibility of a patient going into sepsis.
Early detection and treatment started up to 4 hours before the onset of Sepsis for emergency care patients
Deep Sepsis algorithm has been clinically validated.