Timely monitoring of hemodynamic instability of patient in the ICU is important to prevent any organ dysfunction and mortality. Monitoring hemodynamically unstable patient involves invasive arterial blood pressure & contact-based examination such as center to peripheral temperature difference (CPD), which can be delayed, laborious and infection prone. Thus, we developed a pipeline which make use of thermal imaging based CPD for the prediction of hemodynamic shock. Our models achieved AUROC of 79% and 70% for next 3hr and 12hr forecast respectively. Automated detection achieved 96% agreement with human annotations.
Describe the new knowledge and additional skills the participant will gain after attending your presentation.: Participants will learn about a new modality of shock prediction using thermal imaging. They will get to learn about a computer vision pipeline on infrared signals with open-source codes for implementing the prediction models in their own settings. Our method involve use of an affordable thermal imaging technology and a mobile phone where we have integrated a medical hypothesis with machine learning which delivers decisions in ICU. Thus, participants will learn about development of a decision support system for a clinical setting.
Aditya Nagori (Presenter)
CSIR- Institute of Genomics and Integrative Biology
Lovedeep Singh Dhingra, All India Institute of Medical Sciences
Ambika Bhatnagar, All India Institute of Medical Sciences
Rakesh Lodha, All India Institute of Medical Sciences
Tavpritesh Sethi, All India Institute of Medical Sciences