Imagine if someone could tell whether your loved one has sepsis merely by looking at the palm of their hand? Or at least using a specialized camera to look at the palm of their hand. The key to survival from sepsis is early recognition, so clinicians and researchers are continuously refining techniques to reduce the time it takes to diagnose and treat sepsis. Machine learning is being applied to help to refine what data is most critical in reaching a speedier diagnosis. And some research suggests that using a specialized camera on a sick patient’s hand may shorten the time to diagnose – and treat – sepsis.
Many clinicians have been using a version of a sepsis calculator to help determine if the patient in front of them is showing early signs of sepsis. Many hospitals use an alert system built into their health records that automatically calculates a sepsis risk score based on the patients vital signs. If the patient is at high risk of sepsis, clinicians will be presented with a pop up in the patient’s health records that draw attention to the clinician. This is particularly important in emergency situations where providers are trying to address multiple problems at once. But no system has been proven to be the definitive method for early identification of this deadly disease.
Machine learning has been finding possible relevant data in unexpected places that may yet inform future clinical decision making. Researchers at University Hospital in Heidelberg have aimed their attention, and cameras, at data from patients’ skin. More specifically, they are using spectral imaging data from a camera tuned to wavelengths the human eye cannot see. The scan picks up subtle shifts in blood, oxygen, and water content in the skin, signs of the microcirculatory changes that occur early in sepsis, as small blood vessels become more permeable and leak fluid into surrounding tissue.
Researchers found this imaging helped rapidly diagnose sepsis and predict mortality, giving some hope to future use of similar tools. The study cohort was relatively small – approximately 500 patients, but not too small to ignore. They found that patients with sepsis have significantly lower tissue oxygen saturation and higher tissue hemoglobin and water content than patients without sepsis. When combined with clinical presentation, this hyperspectral imaging (HIS) analysis led to high accuracy in identifying early sepsis.
This is no single magic bullet, however, and more research needs to be done to demonstrate just how reliable, and accessible, HSI is for catching sepsis and saving lives.
