A burn depth detection system based on near infrared spectroscopy and ensemble learning.

Near infrared (NIR) spectroscopy can effectively detect the changes in the burned tissue. However, due to the complex relationship between the spectral signals and the burn depth, simple methods of data analysis are difficult to solve this problem effectively. Therefore, in this paper, a machine learning method is introduced into the NIR spectral signal analysis, which is used to establish the relationship between NIR spectral signals and burn depth. First, based on the intensity of the spectral signal and the diffuse reflection theory, the optical properties that can reflect the change of burned tissue are extracted. And then the chained-agent genetic algorithm (CAGA) optimized support vector regression (SVR) is applied to establish a regression model between the optical property parameters and burn depth. Finally, the porcine model was used for verification. The experimental results demonstrate that the proposed CAGA-SVR integrated inversion model with optical properties can perform accurate inversion of burn depth and provide a reference for doctors.