Title : Innovative sensor for rapid sepsis detection using Raman Spectroscopy
Abstract:
Background: Sepsis being one of the leading causes of mortality and hence causing a burden on the healthcare system without a standard detection test. Therefore, this in turn delays the time to detection leading to non-responsive empirical treatment options. The huge burden of sepsis and the rapid onset requires a rapid culture- free detection technique, Raman spectroscopy along with Surface enhanced silver nanoparticle based substrate can provide a highly sensitive platform. SERS based rapid detection in combination with AI can enhance the detection with high sensitivity and specificity. The AI model for SERS was trained during the study for the ESKAPE pathogens involved in Sepsis. The aim of the study is to optimize, train and validate the AI model for the detection of Sepsis. Large data of eskape pathogens were collected for the test, training and validation of the model. The process of sample collection required optimization for biomarker concentration in the biofluids. The POC validation for clinical adaption is the ultimate objective of the study.
Method: The SERS has been trained on over 1600 samples of different pathogens including bacterial and fungal infections. The samples were collected from the blood culture bottles and centrifuged to remove the cell debris and to concentrate the bacterial biomarkers. The biomarkers will be concentrated in the supernatant, using a very low volume of the sample 2 µl loaded on to the silver nanorod and exposed to the laser at 785 nm for the shooting of the spectra. 30-40 spectra were collected from different spots of the silver nanorods.
Results: The current study shows a very highly sensitive AI model for sepsis detection. The model shows an overall accuracy of 99.09%. The validation of the model can further conclude the efficacy of the model for adaption as a POC model.
Conclusion: Rapid detection of Sepsis using an AI model can serve as a POC for detection in critical care patients.
