Sepsis is a serious condition induced by an infection, often by a bacterial pathogen, leading to organ damage or even death. Despite numerous advances in medicine over the years, the condition still affects millions of people in both developed and developing countries. In the US, sepsis affects 1.7M and kills over 265,000 people annually. Sepsis mortality rates in developing countries are substantially higher. In terms of demographics, sepsis affects humans of all age and race, but it is most pronounced at the age extremes (infants and the elderly) and patients whose immune system is already under strain due to other illnesses or immune system-weakening therapies, e.g., cancer patients undergoing chemotherapy. Blood cultures are currently the default technique used in detecting and diagnosing the root cause of sepsis. However, blood-cultures can take upwards of 24-48 hours in order to obtain results. In that time, the patient can experience irreversible harm due to the condition if not treated properly. Unfortunately, precise and effective antibiotic treatment requires knowledge of the pathogen causing sepsis. Beyond a long time to get an answer, blood cultures often exhibit alarmingly high false negatives (failure to detect a pathogen causing sepsis) and typically do not precisely identify the pathogen causing sepsis. Hence there have been several efforts aimed at detecting and identifying the broad range of potential pathogens causing sepsis and circumventing the need for blood cultures. However, many of the recently proposed methods for detecting and diagnosing sepsis exhibit one or more of the following drawbacks: (i) they lack high sensitivity (ability to detect a pathogen); (ii) they cannot accurately identify a broad range of pathogens from a single sam- ple; (iii) take a (relatively) long time; (iv) require a large volume of blood; or (v) cannot be used in the real-time monitoring of sepsis (either detecting pathogens known to cause sepsis or quantifying the patient's response to antimicrobial treatment). We are proposing a new sepsis detection method, combining ?ow imaging microscopy (a high-throughput technique for imaging millions of microscopic particles) and deep learning based image analysis (techniques leveraged in facial recognition and self-driving cars) to overcome the above mentioned limitations. The approach has proven capable of detecting a variety of bacterial species in low concentrations of mouse blood in less than 1 hour of processing time with as little as 50 L of blood. In this proposal, one of our aims is to optimize our approach and quantify the accuracy and limits of detection in human blood. Our patent pending approach has also been licensed to a major manufacturer of ?ow imaging microscopes. Another aim of this research is to begin integration of our technology with an existing commercial instrument with the intention of providing a compact self- contained device that can be deployed at numerous hospitals world-wide. The implementation of our platform should have a major impact on antimicrobial treatment in all areas of the hospital.
Sepsis affects 1.7M US citizens (causing roughly 270k deaths) each year; the condition is also the most expensive condition treated in US hospitals costing approximately 24B USD each year. Current methods for detecting and determining the source of the infection causing sepsis are inaccurate, too slow, and do not provide detailed pathogen speci?c information needed for effective treatment. Our proposal aims at developing a fast approach, combining ?ow imaging microscopy and deep learning, for detecting and determining the root cause of sepsis from blood samples (addressing many issues facing sepsis detection and diagnosis) which can deployed at a variety of hospitals worldwide.
