Project

AI for Early Detection of Sepsis & Infections

The Need
Early detection of sepsis is a highly recognized need in healthcare to reduce mortality and improve patient outcomes.

Our Approach

Our team of researchers in collaboration with doctors at Amrita Institute of Medical Sciences is actively looking at ways to predict onset of sepsis ahead of time so as to save lives. We work on the following three major goals.

1. Use data driven techniques to understand the physiological presentations of sepsis and thereby use sensing devices generated data to more precisely identify early signs of sepsis
2. Impact of IoT connected remote patient monitoring system for patients with high risk of rapid deterioration and its integration to the rapid response team in preventing Out-of-ICU ward crashes
3. Development of explainable-AI tool for optimizing antibiotic prescriptions in patients with hematological malignancies

Associated Funders

1. ICMR 2021-12140 grant funds “An observational prospective study to assess the impact of Internet-of-Things connected remote patient monitoring system for patients with high risk of rapid deterioration and its integration to the rapid response team in preventing Out-of-ICU ward crashes”.

Research Collaborators

Amrita Institute of Medical Sciences, Kochi, India (https://www.amrita.edu/school/medicine/kochi/) Stanford Healthcare Innovations Labs at Stanford Medical School, US (https://innovations.stanford.edu/home)

Publications

1. Ekanath Srihari Rangan, Rahul Krishnan Pathinarupothi, Kanwaljeet J S Anand, Michael P Snyder, Performance effectiveness of vital parameter combinations for early warning of sepsis—an exhaustive study using machine learning, JAMIA Open, Volume 5, Issue 4, December 2022, ooac080, https://doi.org/10.1093/jamiaopen/ooac080
2. Pathinarupothi, Rahul Krishnan, Dipu T. Sathyapalan, Merlin Moni, KA Unnikrishna Menon, and Maneesha Vinodini Ramesh. “REWOC: Remote Early Warning of Out-of-ICU Crashes in COVID Care Areas using IoT Device.” In 2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 2010-2013. IEEE, 2021.
3. Shaji, S., Pathinarupothi, R.K., Rangan, E.S., Menon, K.U. and Ramesh, M.V., 2021, October. Heart Lung Health Monitor: Remote At-Home Patient Surveillance for Pandemic Management. In 2021 IEEE Global Humanitarian Technology Conference (GHTC) (pp. 127-130). IEEE.

Tools

Vital-SEP is a Sepsis Prediction Engine that employs Gradient Boosted Decision Tree (XGBoost) on features extracted from vitals obtained from wearable sensors. We have open sourced our implementation as well as the models for the larger research community. https://pprahul.github.io/Vital-SEP/

People Involved

Dr. Unnikrishna Menon

Dr. Rahul Krishnan

Dr. Dipu T S – https://amritahospitals.org/team/dr-dipu-t-s/ 

Dr. Merlin Moni – https://amritahospitals.org/team/dr-merlin-moni/