No abstract

BACKGROUND Currently, hyperspectral technology has been used in various fields, but its applications for the detection of chylous plasma are lacking. This paper used hyperspectral techniques in combination with machine learning algorithms for the detection of chylous plasma, providing a new diagnostic method. OBJECTIVE This paper proposed a method of plasma chylous degree detection and recognition based on machine learning and hyperspectral technology. A plasma chylous degree detection model was established.It fills the gap of machine learning and hyperspectral technology in the detection of chylous plasma. METHODS The plasma hyperspectral data were preprocessed using the multiple scattering correction (MSC) method and then classified using four classification algorithms, including random forest (RF), K-nearest neighbor KNN), Perceptron, and stochastic gradient descent (SGD) algorithms and the best algorithm was compared.Finally, band selection is carried out to screen the best band subset. RESULTS The results showed that the random forest algorithm had the best effect. Then, the model of plasma chylous degree detection based on random forest was established. Finally, 10 important spectral bands, including 1192.45 nm, 1182.9 nm, 946.98 nm, 1202.01 nm, 1080.93 nm, 1278.41 nm, 1237.03 nm, 991.65 nm, 1020.35 nm, and 1697.8 nm, were selected by band selection. After adjusting the parameters to optimize the model, the prediction accuracy of the whole band was 0.89. CONCLUSIONS This study suggested that hyperspectral technology could identify chylous plasma and could be used to improve its detection efficiency in biomedicine, human function tests, and other aspects.

Blood transfusion is a critical medical treatment, which is performed to save patients’ lives. Chylous blood had high fats. The transfusion of chylous blood into a patient can cause the blockage of micro-vessels. Most blood collection stations are not equipped with the equipment for the detection of chylous blood, and the detection is usually performed with direct observation through the human naked eye, which is prone to certain human errors. Only a few large blood collection stations use the equipment for the detection of chylous blood. In this study, plasma hyperspectral data were collected to detect and identify chylous plasma. The data were preprocessed using the multiple scattering correction (MSC) method and then classified using four classification algorithms, including random forest (RF), K-nearest neighbor KNN), Perceptron, and stochastic gradient descent (SGD) algorithms. First, the healthy and chylous plasma samples were classified into simple dichotomies. The best algorithm was identified by comparing the results of classification algorithms. The results showed that the random forest algorithm-based classification model had the best effect.Then, the chylous plasma was subdivided into different degrees of chylous plasma, which were less separable.A random forest algorithm-based plasma chylous degree detection model was established. Finally, 10 important spectral bands, including 1192.45 nm, 1182.9 nm, 946.98 nm, 1202.01 nm, 1080.93 nm, 1278.41 nm, 1237.03 nm, 991.65 nm, 1020.35 nm, and 1697.8 nm, were selected by band selection. After adjusting the parameters to optimize the model, the prediction accuracy of the whole band was 0.89. This study suggested that hyperspectral technology could identify chylous plasma and could be used to improve its detection efficiency in biomedicine, blood donation centers, human function tests, and other aspects. Filling the gap between machine learning and hyperspectral technology.To provide a new method for the diagnosis of chylous plasma.

Traditional hyperspectral image semantic segmentation algorithms can not fully utilize the spatial information or realize efficient segmentation with less sample data. In order to solve the above problems, a U-shaped hyperspectral semantic segmentation model (DCCaps-UNet) based on the depthwise separable and conditional convolution capsule network was proposed in this study. The whole network is an encoding–decoding structure. In the encoding part, image features are firstly fully extracted and fused. In the decoding part, images are then reconstructed by upsampling. In the encoding part, a dilated convolutional capsule block is proposed to fully acquire spatial information and deep features and reduce the calculation cost of dynamic routes using a conditional sliding window. A depthwise separable block is constructed to replace the common convolution layer in the traditional capsule network and efficiently reduce network parameters. After principal component analysis (PCA) dimension reduction and patch preprocessing, the proposed model was experimentally tested with Indian Pines and Pavia University public hyperspectral image datasets. The obtained segmentation results of various ground objects were analyzed and compared with those obtained with other semantic segmentation models. The proposed model performed better than other semantic segmentation methods and achieved higher segmentation accuracy with the same samples. Dice coefficients reached 0.9989 and 0.9999. The OA value can reach 99.92% and 100%, respectively, thus, verifying the effectiveness of the proposed model.