Design Of The Material Discrimination And Display System Based On Multi-energy Spectrum X-ray Imaging

X-rays have strong penetrability,and X-ray imaging can be realized by taking advantage of the different attenuation capabilities of various organs and tissues of the human body.X-ray imaging is fast,low cost,and practical,so it is widely used in medical imaging and auxiliary medical diagnosis.With the continuous improvement of the counting rate and counting accuracy of photon counting detectors,the material discrimination and display system based on multi-energy spectrum X-ray imaging is gradually entering the field of clinical medical diagnosis.In this thesis,a multi-energy spectrum X-ray source and linear array detector are used to construct a multi-energy spectrum X-ray material resolution and color display demonstration system,which not only realizes the resolution of several common human organs and tissues through the classification model,but also distinguishes the material The result is combined with the photon count value to draw a color image of the measured object.From the color image,the properties and thickness of the substance can be distinguished intuitively.Experiments have proved that this system has high material resolution accuracy and high quality color images.In terms of material discrimination,this thesis uses the XGBoost multi-classification model,taking 4 phantoms of common human organs and tissues as the object to be measured,and taking the photon count value and material attribute R of the phantom under 5 energy thresholds as the characteristic values.With the phantom type as the target value,an XGBoost multi-classification model was trained,and the hyperparameters of the model were tuned by grid search and cross-validation methods,and the material discrimination effect of the model was evaluated from multiple angles.Experiments show that the model has a high material resolution accuracy and can effectively distinguish four types of organs and tissues.In terms of imaging display,this thesis uses the HSI color model as the basic imaging model,and maps the XGBoost material classification results and the photon count values of the five energy thresholds of the measured object to the chromaticity,saturation and brightness of the HSI model to achieve the Color display of the measured object.Experiments have proved that the image imaging effect of color display is better,and the type of substance and thickness information can be distinguished from the image.