The Burn Depth Detection System Based On Near Infrared Spectrum And Ensemble Learning

There are many burn patients in the world every year,and the detectionof burn depth is important for burn diagnosis.At present,the clinical diagnosis of burn depth depends on the experience of clinicians,but even for the most experienced clinicians,the accuracy of the depth classification can only reach 65%-70%.This inaccurate diagnosis can have serious consequences,such as unnecessary infection,pain and even death.Near infrared spectroscopy(NIR)is a non-invasive,non-contact and rapid detection method,which is very suitable for the detection of burn depth.However,few research reports on how to analyze the near infrared spectrum data to obtain the depth of burn,which greatly limits the application of near infrared spectroscopy in the detection of burn depth.For the above existing problem in the burn depth dection,in order to promote the application of near infrared spectroscopy in the detection of burn depth,and lay the foundation of its theory and methods.This thesis proposed the burn depth detection systems of near infrared spectral data based on ensemble learning.The main contents are as follows:(1)By using the light intensity signal of near infrared spectrum,this thesis proposed the integrated regression model of support vector regression with the chain-like agent genetic algorithm(CAGA)(CAGA-SVR)and random forest(RF)regression model respectively,so as to form two kinds of burn depth detection system.First,use fiber optic spectrometer to collect NIR spectrum,and support vector regression(SVR)is used to generate the burn depth detection model.Then,optimize the SVR model by CAGA to achieve the CAGA-SVR burn depth detection system.In addition,based on the near infrared spectral signal,a RF burn depth detection system was also constructed.(2)By collecting the spectrum signal based on near infrared spectrometer system,extract optical properties,which can reflect the cell structure and composition of burn tissue.The parameters was used with CAGA-SVM to construct a burn depth detection system which was an CAGA-SVR ensemble learning.First,collecte the near infrared spectrum and extract the optical properties through the diffuse theory.Then,constrct a burn depth detection model combing the parameters with SVR.Finally,optimize the model by CAGA to improve the performance.(3)By collecting the spectral image signal based on the spectral imagingsystem,extract optical property parameters,which can reflect the cell structure and composition of burn tissue.This thesis proposed an integrated burn depth detection system based on spectral image signal and RF.Firstly,the spectral image signal is acquired.Then the optical properties were extracted by diffuse theory,the correlation between the parameters and the burn depth was also analyzed.Combined the light intensity signal with RF,a burn depth detection system based on spectral images was constructed.