Study On Detection Of Wound Infection Based On Near-infrared Spectroscopy

In the field of clinical trauma surgery,bacterial infection is a very important complication,which will affect the therapeutic effect.The diagnosis of wound infection can prevent further deterioration of the wound and relieve the patient’s pain.Currently clinical diagnosis of bacterial infection relies mainly on the specific microbiological and biochemical identification.For these methods,bacterial culture and colony counting will take a lot of time and effort,and biopsy involves sections,staining,and microscopic evaluation,which are invasive and time-consuming.Therefore,there is an urgent need to establish an accessible,simple,noninvasive and effective method for the clinical detection of wound bacterial infection.As a noninvasive,non-contact and rapid detection method,near infrared spectroscopy is very suitable for detection of wound bacteria infection.However,for the previous near infrared technologies the extract features are not physical meaningful and they were not applied for detection of wound infection in vivo.It limited the understanding of the changes in the structure and composition of the infected wound and restricted the application of clinical bacteria detection.To solve these problems,this thesis proposed a optical detection method of wound infection based on near infrared spectroscopy combined with optical properties.it aims to further study the application of nearinfrared spectroscopy in the detection of wound infectionand to analyze the wound bacteria infection from the physical layer.The main research work of this thesis is listed as follows:Aiming at the problem of information detection of wound bacterial solution,a classification model based on optical properties combined with support vector machine(SVM)was proposed.Light intensity signals were collected from seven common wound bacteria solutions with the same concentration.The optical properties(reduced scattering coefficient(?’_s)and absorption coefficient(?_a))were extracted from the spectrum as the features.Then,based on the features,the classification model of SVM and SVM optimized by particle swarm optimization(PSO-SVM)were used to identify seven kinds of bacteria with the same concentration.SVM was used to classify two kinds of bacteria under different concentrations based on the optical properties.Aiming at the problem of wound bacteria detection in vivo,two detection methods based on optical properties were proposed,which correspond to two kinds of data acquisition methods.For the fiber spectrometer,the method for detecting bacteria in vivo was proposed based on optical properties and integrated classification model.The spectral signals were obtained by using near infrared fiber spectrometer,then,optical properties which can reflect the internal structure and composition of the wound are extracted based on spectral signals.Based on the optical properties,the SVM optimized by chain-like agent genetic algorithm(CAGA)(CAGA-SVM)classification model is used to identify the wound bacterial infection in vivo.For the hyperspectral imagingsystem,the detection method of wound bacteria based on the optical properties was proposed.The spectral signal of each pixel is obtained by the imagingsystem.Optical properties are extrated by diffusion theory,then,the distribution maps of optical properties are obtained.By analyzing the correlation between the optical properties and the wound bacterial infection in vivo,it can demonstrate the feasibility of the optical properties for full-field detection of infected woundThis thesis provides the thought and realization method of detection of wound infection based on physical nature of wounds.It provides in-depth theoretical basis for the detection wound bacterial infection based on near infrared spectroscopy combined with optical properties,which has a certain theoretical significance and practical significance.